Thermodynamics and Kinetics of Ceramic/Metal Interfacial Interactions

Thermodynamics and Kinetics of Ceramic/Metal
Interfacial Interactions
by
Raymundo Arryave
o
B.S., Instituto Tecnolgico y de Estudios Superiores de Monterrey
o
(1996)
S.M., Massachusetts Institute of Technology (2000)
Submitted to the Department of Materials Science and Engineering
in partial fulllment of the requirements for the degree of
Doctor of Philosophy in Materials Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
February 2004
c Massachusetts Institute of Technology 2004. All rights reserved.
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Materials Science and Engineering
November 25, 2003
Certied by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thomas W. Eagar
Thomas Lord Professor of Materials Engineering and Engineering
Systems
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Harry S. Tuller
Professor of Ceramics and Electronic Materials
Chairman, Department Committee on Graduate Students
Thermodynamics and Kinetics of Ceramic/Metal Interfacial

Interactions
by
Raymundo Arryave
o
Submitted to the Department of Materials Science and Engineering
on November 25, 2003, in partial fulllment of the
requirements for the degree of
Doctor of Philosophy in Materials Science
Abstract
Ceramic/metal interfaces occur in a great number of important applications, such
as ceramic/metal composites, microelectronics packaging, ceramic/metal seals, and
so forth. Understanding the formation and evolution of such interfaces is therefore
essential for the better design and optimization of these technologies.
In this thesis, a methodology for the study of the thermochemical interactions
at ceramic/metal interfaces, during both their formation and evolution, is proposed.
Because of the importance of zirconia-based ceramics in increasingly important applications such as structural composites, thermal barrier coatings and Solid Oxide Fuel
Cells, it was decided to illustrate the concepts developed in this thesis through the
study of the interactions between zirconias and active metals.
Semi-empirical thermodynamic models of all the phases likely to take part in
the ceramic/metal interfacial interactions studied were developed. Phase diagram
data and thermochemical information were critically assessed and use to adjust the
thermodynamic parameters that allowed the description of the Ag-Cu-Ti, Cu-Ti-Zr,
Ti-Zr-O, Cu-Ti-O and Cu-Zr-O systems.
The thermodynamic models were used to predict the diusion paths across zirconia/active metal interfaces through metastable phase diagrams calculations. Additionally, equilibrium calculations of activity diagrams were used to understand the
complex interfacial reactions occurring during the active metal brazing of zirconiabased ceramics. By using simple one-dimensional inter-diusion simulations, it was
demonstrated that the base metal in ceramic/metal joints plays an essential role
in determine the thermochemical interactions at the ceramic/metal interface during
ceramic/metal joining operations. In general it was found that, using all these techniques, it was possible to explain diusion paths and reaction sequences observed in
a great number of zirconia/active-metal systems, both in the solid and in the liquid
states.
In many cases, the morphology of the reaction layers formed at ceramic/metal
interfaces determine their nal properties. To address this problem, empirical ther3
modynamic models of the likely reaction products at zirconia/metal interfaces were

coupled to kinetic models using the diuse-interface formalism to successfully describe
the formation and evolution of complex ceramic/metal interfacial structures.
Thesis Supervisor: Thomas W. Eagar
Title: Thomas Lord Professor of Materials Engineering and Engineering Systems
Acknowledgments
This thesis is dedicated to my parents, Graciela and Raymundo Arryave. I could
o
not have come all the way to this stage in my academic life if it not were because of
their encouragement, support and love, Gracias Paps!.
a
First of all, I would like to thank my advisor, Prof. Thomas W. Eagar. I cannot
think of any other advisor from whom I could have learned as much as I did. He was
not only available when I needed counsel regarding the direction of my research, but
he was also there when I needed career and even personal advise. I deeply admire the
respect that he has for all his students, and for the fact that, for him, our personal
well-being is as important (or even more important, indeed) as our performance as
researchers. He is a great human being, and I hope that I not only learned from
his vast knowledge in the eld of Materials Science but also from his even greater
knowledge of life and human character. I dont know if this happens very often at
MIT, but Prof. Eagar took the risk of giving me freedom to pursue the research path
that best tted my interests and goals and for that Im deeply thankful. I hope that
this work is up to his expectations.
I would also like to thank the other members of my committee. I thank Dr.
Larry Kaufman for all the help that he gave me when I was starting to explore thermodynamic modeling as a possible way to approach the problem of ceramic/metal
interactions. I consider myself very fortunate to have had the opportunity to interact
with one of the pioneers in the eld of computational thermodynamics, but I consider
my self even more fortunate and honored for his friendship. I would also like to thank
one of the great ones in Thermodynamics of Materials, Prof. Claude Lupis. Ive read
his book I dont know how many times, and I still learn new concepts every time I
read it. More important than his vast knowledge is the fact that he is, without a
doubt, one of the nest persons that I met at MIT. I would also like to thank Prof.
Carter. I had the opportunity of being his TA for 3.00 and it was there that I learned
how important it is to enjoy the subject (Thermodynamics) that you teach. All I
have left to say is that it has been a great honor to have all these great material
scientists and more important, persons, as members of my committee.
I could not have done this without the help and support from all the members of
the Welding and Joining Group. I would like to thank Don Galler for all the advise
and help he provided while seting up my experimental apparatus. I thank Jeri for
all her patience and for all the interesting conversations about movies that we had
while waiting for my meetings with Tom. I would like to thank Chris Musso for his
friendship and all his help in designing my experimental aparatus, and also for all
the great discussions (sometimes a tad heated, but always respectful) about so many
topics. Neil: you are a weird guy...just kidding, it was great to meet you and is an
honor to be considered your friend. Joel: thank you for all the drawings and all
your help with my experiments. I would also like to thank Harold for all his help
and support. Finally, I would like to thank my old labmates Mike, Burke, Patricio,
5
Jocelyn and Jin-Woo, for their friendship and help.

Life at MIT can be sometimes a little bit rough (no sarcasm here). To keep
sanity, everyone must have his/her support group of equally stressed out, depressed,
confused, students. I was very fortunate to have met so many smart and great people in my particular support group. I thank my friend, cousin, former roommate
and even former labmate Chorro for his help and support during my rst years here
at MIT. Jorge Carretero: weve been through regular, bad, and even worse times.
Thanks for your friendship. I would also like to thank Jorge Feuchtwanger, not only
a good friend, but also one of the few persons that did not have to read this thesis,
and despite this, he read it and gave me valuable advise! I also thank my favorite
roommie, Gina, for her friendship and support. I thank the rest of the Mexican Gang:
Ante, Juan Gonzalez, Ulises, Oso, Antonio, Nuria, Lula. I want to thank especially
Kate Baty, who is an angel that worked uder the cover of Director of The Host Families Program for International Students at MIT. She adopted me in the great and
numerous Baty Family and gave me something that is always missed: the feeling of
being home. Finally, I would also like to thank Gerardo Trpaga, who, during my
a
rst two years at MIT provided me with invaluable advise and more importantly,
with his frienship.
Back in Monterrey Mexico, there are many people that I want to thank. First of
all, I thank my Mom and Dad for their unconditional love and support. I also thank
Grace and Jorge for being the coolest sister and brother one can have. Bere: I love
you and I thank you for being so supportive during the past couple of long, crazy
months. I deeply thank my Aunt Nene, since it was thanks to her that I had nancial
support during my undergraduate years at ITESM. Without her support, I might not
have been able to come here. Abuelita: thanks for all your prayers and your love. I
thank all my uncles, aunts, cousins, nephews and nieces back in Monterrey for their
moral support. I thank also all my friends in Monterrey. They have helped me keep
my feet on the ground. To them Im only Ray, and I hope that I will always remain
that way.
Contents
1 Introduction
21
2 Ceramic/Metal Interfaces: Basic Ideas, Joining Processes

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Wetting as a Metric of Ceramic/Metal Interfacial Energetics . . . . .
2.3 Reactive vs. Non-Reactive Wetting . . . . . . . . . . . . . . . . . . .
2.3.1 Ceramic Dissolution: Non-reactive Case . . . . . . . . . . . .
2.3.2 Addition of Reactive Elements . . . . . . . . . . . . . . . . . .
2.4 Non-reactive Metal-Ceramic Wetting: Inuence of the Electronic Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 Joining Ceramics to Metals . . . . . . . . . . . . . . . . . . . . . . .
2.6 Active Brazing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7 Understanding Ceramic/Metal Interfacial Interactions: The need for
Thermodynamic Databases. . . . . . . . . . . . . . . . . . . . . . . .
25
25
25
27
27
29
31
34
36
39
3 Development of Thermodynamic Database for Active-Brazing Alloys

3.1 Development of Active Brazing Thermodynamic Database: Introduction
3.2 Assessment of the Ag Cu T i System . . . . . . . . . . . . . . . .
3.2.1 Assessment of the Ag T i Subsystem . . . . . . . . . . . . .
3.2.2 Assessment of the Ag Cu T i Ternary . . . . . . . . . . . .
3.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Assessment of the Cu T i Zr System . . . . . . . . . . . . . . . .
3.3.1 Description of the Cu T i Zr System . . . . . . . . . . . .
3.3.2 Phases Described in the Cu T i Zr System . . . . . . . . .
3.3.3 Optimization of the Cu T i Zr System: Results . . . . . .
3.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
41
42
43
50
54
55
55
56
57
59
4 Development of Thermodynamic Database for Ceramic Oxide Substrates and Reactive Brazing Interfacial Products
4.1 CALPHAD Modelling of Ceramic Oxide Systems: Introduction . . .
4.2 Assessment of the Zr O System . . . . . . . . . . . . . . . . . . . .
4.2.1 Phases present in the Zr O System . . . . . . . . . . . . . .
4.2.2 Model Optimization and Comparison with Experimental Results
4.2.3 Calculated Zr O phase diagram. . . . . . . . . . . . . . . .
61
61
62
63
65
68
4.3
4.4
4.5
4.2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modelling of the Cu T i O System . . . . . . . . . . . . . . . . .
4.3.1 Experimental Phase Diagram of the Cu T i O System. . .
4.3.2 Thermal Stability range of M6 X compounds . . . . . . . . . .
4.3.3 Determination of Thermochemical Properties of M6 XCompounds
4.3.4 Modelling of the M6 X Compounds and Phase Diagram Calculation for the Cu T i O System . . . . . . . . . . . . . . .
4.3.5 Modelling the Solubility of Al in M6 X Compounds in the Cu
T i O System. . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Critical assessment of the Cu Zr O liquid phase. . . . . . . . . .
4.4.1 Cu Zr O Melts . . . . . . . . . . . . . . . . . . . . . . . .
4.4.2 Assessment of the Validity of the Thermodynamic Model for
Cu O Zr melts. . . . . . . . . . . . . . . . . . . . . . . .
4.4.3 Wagner interpolation method. . . . . . . . . . . . . . . . . . .
4.4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Thermodynamic Assessment of the T i Zr O System . . . . . . .
4.5.1 Phases Present in the T i Zr O System . . . . . . . . . . .
4.5.2 Optimization and Comparison to Experimental Data I: 14500 C
Isothermal Section . . . . . . . . . . . . . . . . . . . . . . . .
4.5.3 Optimization and Comparison to Experimental Data II: Liquidus Surface . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5.4 Optimization and Comparison to Experimental Data III: T iO2
ZrO2 pseudobinary section . . . . . . . . . . . . . . . . . . . .
4.5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
69
69
69
70
74
74
76
76
76
79
80
82
82
82
84
86
89
91
5 Application of Thermodynamics to the Study of Ceramic/Metal Interfaces

93
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Interfacial Interactions between Zirconia and Reactive Metals: Motivation and Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.2.1 Motivation for the Study of Zirconia/Metal Interactions . . . . 94
5.2.2 Interactions between Zirconia and Active Elements . . . . . . 95
5.3 Thermochemistry of Ceramic-Metal Interfaces: Previous Work . . . . 98
5.3.1 Thermochemistry of Active Metal Brazing Systems . . . . . . 99
5.3.2 Application of Chemical Potential and Phase Diagrams to the
Study of Ceramic/Metal Interactions . . . . . . . . . . . . . . 99
5.3.3 Diusion Simulations of T i/Al2 O3 Metastable Diusion Couples 100
5.3.4 Use of Activity Diagrams: A More General Approach . . . . . 100
5.3.5 Limitations and Application to Zirconia/Metal Interactions . . 101
5.4 Application of Thermodynamic Models to the Study of Zirconia/Ti
Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4.1 Interfacial Reactions between Liquid T i and ZrO2 . . . . . . . 102
5.4.2 Interfacial Reactions between Solid T i and ZrO2 . . . . . . . 111
5.4.3 Comparison with Al2 O3 /T i Interactions . . . . . . . . . . . . 117
8
5.5
5.6
5.7
5.8
Thermodynamic Study of the Interactions between Zirconia and Ag

Cu T i Brazing Alloys for Ceramic/Metal Joining Applications . . .
5.5.1 Thermochemistry of Ag Cu T i Alloys . . . . . . . . . . .
5.5.2 Interfacial Reactions Occurring during Ag Cu T i Brazing
of Zirconia . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.3 Thermodynamic Analysis of ZrO2 /Ag Cu T i/T i Brazed
Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.4 Thermodynamic Analysis of ZrO2 /Ag Cu T i/ZrO2 Brazed
Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interfacial Reactions in the Zirconia/Cu-Ti-Al/Ni System . . . . . . .
5.6.1 Experimental Results . . . . . . . . . . . . . . . . . . . . . . .
5.6.2 Analysis of Experimental Results . . . . . . . . . . . . . . . .
Interaction between Cu Zr melts and Zirconia . . . . . . . . . . . .
Application of Thermodynamics to the Study of Zirconia/Metal Interfaces: Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
118
119
121
123
125
126
127
130
134
136
6 Metal Substrate Eects on the Thermochemistry of Active Brazing

Interfaces
139
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.2 Previous Experimental Evidence . . . . . . . . . . . . . . . . . . . . . 140
6.3 Substrate-Brazing Alloy Interaction: Simplied Analysis . . . . . . . 141
6.4 Thermochemical Analysis of Metastable Interfacial Systems . . . . . . 142
6.5 Kinetic Analysis of liquid/f cc Moving Boundary. . . . . . . . . . . . 143
6.6 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.7 Experimental Verication: Zirconia/Cu-Ti/Ni Interactions . . . . . . 150
6.8 Metal Substrate Eects in Ceramic-Metal Joining: Concluding Remarks152
7 Using Phase Field Techniques to Model Coupled Oxide Growth during Active Brazing of Ceramic Oxides
155
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.2 Interactions between Zirconias Ceramics and Ti-containing Brazes . . 156
7.3 Current Modelling of Coupled-Oxide Growth . . . . . . . . . . . . . . 159
7.4 Thermodynamics and Kinetics of Heterogeneous Systems . . . . . . . 161
7.5 Gibbs Energy Expression to model Coupled Growth System . . . . . 165
7.6 1D Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.6.1 Numerical Simulations: Inuence of k Parameter . . . . . . . 170
7.6.2 Analysis of Numerical Results . . . . . . . . . . . . . . . . . . 172
7.7 2D Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . 178
7.8 Experimental Verication of Phase-Field Model . . . . . . . . . . . . 180
7.8.1 Examining the Local Equilibrium Hypothesis . . . . . . . . . 181
7.8.2 Experimental Verication of Diusion-Controlled Kinetics . . 182
7.8.3 Experimental Evidence for the Stability of Planar vs. Undulated Interfaces in the T i O System . . . . . . . . . . . . . . 183
7.9 Sensitivity of the Model To the Thermodynamic Description . . . . . 186
7.10 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . 187
9
7.10.1 Conclusions for this Chapter . . . . . . . . . . . . . . . . . . . 187

7.10.2 Future Improvements on the Model . . . . . . . . . . . . . . . 189
8 Conclusions and Future Work
191
8.1 General Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
A The Calphad Approach to Thermodynamic Modelling: Developing
models for Ceramic/Metal Systems
195
A.1 The CALPHAD Methodology: An Introduction . . . . . . . . . . . . 195
A.2 Calphad Models: A Brief Description . . . . . . . . . . . . . . . . . . 196
A.2.1 Stoichiometric Compounds . . . . . . . . . . . . . . . . . . . . 197
A.2.2 Random Substitutional Phases . . . . . . . . . . . . . . . . . . 197
A.2.3 The Sublattice Model . . . . . . . . . . . . . . . . . . . . . . . 199
A.2.4 Applications of the Sublattice Model . . . . . . . . . . . . . . 203
B Derivation of an expression for the Chemical Potential of an element
i in a Sublattice Phase
207
C Model Parameters
209
D Qualitative Assessment of Diusion Paths in AgCu-Ti Couples

223
D.1 Estimation of Diusion Paths . . . . . . . . . . . . . . . . . . . . . . 223
D.2 Application to the AgCu-Ti Diusion Couple . . . . . . . . . . . . . . 225
D.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
E Numerical Implementation of Phase Field Model of
Layers Growth
E.1 Finite Dierence Formulation . . . . . . . . . . . . .
E.2 Treatment of Boundary Conditions . . . . . . . . . .
E.3 Value of the gradient energy term . . . . . . . . . . .
E.4 Numerical Scheme: Explicit Euler . . . . . . . . . . .
E.5 Numerical Scheme: Semi-implicit Crank-Nicholson . .
E.6 Newton Method . . . . . . . . . . . . . . . . . . . . .
E.7 Implementation of Semi-Implicit Numerical Scheme .
F Experimental Apparatus
Coupled Oxide
229
. . . . . . . . . 229
. . . . . . . . . 229
. . . . . . . . . 230
. . . . . . . . . 230
. . . . . . . . . 231
. . . . . . . . . 232
. . . . . . . . . 233
235
10
List of Figures
2-1 Schematics of contact angle v.s oxide dissolution. . . . . . . . . . . .
2-2 Eects of active element additions on the dissolution of ceramic oxide
substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2-3 Wad v.s electronic properties at C/M interfaces. . . . . . . . . . . . .
2-4 Eg v.s. GOxide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
f
28
3-1
3-2
3-3
3-4
3-5
3-6
3-7
46
49
52
53
54
57
59
Experimental and calculated phase boundaries in Ag T i system. . .

Calculated AgT i phase diagram. . . . . . . . . . . . . . . . . . . . . .
Calculated Ag Cu and Cu T i phase diagrams. . . . . . . . . . . .
Calculated Ag Cu T i 7000 C and 10000 C ternary sections. . . . .
T i chemical activity in Ag Cu eutectic melt at 10000 C. . . . . . . .
Experimental and calculated Cu T i Zr section at 7030 C. . . . . .
Calculated and experimental pseudo-binary section CuT i2 CuZr2 . .
4-1 Experimental vs. calculated Zr O phase diagram. . . . . . . . . . .

4-2 Calculation and experiments of thermochemical properties of solid phases
in the Zr O system. . . . . . . . . . . . . . . . . . . . . . . . . . .
4-3 Calculated and experimental Gibbs free energy of formation of ZrO2x .
4-4 Calculated Zr O phase diagram . . . . . . . . . . . . . . . . . . . .
4-5 Calculated T i and O activities in T i O system at 9450 C. . . . . . .
4-6 Calculated T i and Cu activities in the Cu T i system at 9450 C . .
4-7 Calculated Cu T i O system at 9450 C. . . . . . . . . . . . . . . .
4-8 Eect of the 0 L parameter on the oxygen chemical potential (J/mol)
in a copper solution at 1473 K. . . . . . . . . . . . . . . . . . . . . .
4-9 Eect of the 0 LCu+1 ,Zr+4 :O2 ,V a parameter on the value for Zr in Cu
O
O Zr melts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4-10 Calculated Zr in Cu Zr O liquids using the two-sublattice model
O
and Wagner interpolation at 1473 K . . . . . . . . . . . . . . . . . . .
4-11 Calculated binary phase diagrams belonging to the T i Zr O system.
4-12 calculated and experimental T i Zr O isothermal section at 14500 C
4-13 Calculated liquidus surface for the T i O Zr ternary system. . . .
4-14 Site fraction of T i ions in the ionic liquid phase. . . . . . . . . . . . .
4-15 Comparison between calculated and experimental T iO2 ZrO2 section.
30
32
34
65
66
67
68
71
72
75
80
80
81
85
85
88
90
90
5-1 Calculated 17000 C isothermal section of the T i Zr O system. . . 105

11
5-2 Calculated metastable 17000 C isothermal section of the T i Zr O

system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5-3 Normalized driving forces for precipitation along metastable solid/liquid
phase boundary at 17000 C. . . . . . . . . . . . . . . . . . . . . . . . 108
5-4 Calculated metastable 17000 C ternary section for the T i Zr O
system with ZrO2x , hcp and liquid phases. . . . . . . . . . . . . . 108
5-5 Normalized driving forces for precipitation along the (ZrO2x +hcp)/hcp
phase boundary at 17000 C. . . . . . . . . . . . . . . . . . . . . . . . 109
5-6 Zr O phase diagram, ZrO2 region. . . . . . . . . . . . . . . . . . . 111
5-7 Analysis of zirconia/solid titanium interactions. . . . . . . . . . . . . 113
5-8 Metastable thermodynamic calculation for the T i Zr O system at
14940 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5-9 Diusion paths in the solid T i/ZrO2 system. . . . . . . . . . . . . . . 115
5-10 Zirconia/N i-based super alloy joint using T i thin lm and N i-based
brazing alloy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5-11 Qualitative diusion path in the Al2 O3 /T i system at T 11000 C. . . 118
5-12 Activity of T i in a Ag Cu T i alloy at 11000 C . . . . . . . . . . . 120
5-13 Phase fraction calculation for Cu 40Ag 5T i wt.% active brazing alloy.120
5-14 ZrO2 (Zr) + 2(O) dissolution in Ag Cu T i brazing alloys at
11000 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5-15 Activity diagram for the Cu T i O system at 9000 C . . . . . . . . 124
5-16 Zirconia/Metal brazing using Cu 10 wt.% at 10250 C. . . . . . . . . 127
5-17 Brazing experiments using Cu 5 wt.% brazing alloys. . . . . . . . . 128
5-18 Zirconia/Inconel r joint. Cu 10T i 5Al wt.% braze at 9750 C. . . 129
5-19 High magnication of zirconia/Cu T i Al interface. . . . . . . . . 130
5-20 T i activities of the Cu 10 and Cu 5 brazing alloys as a function of
temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5-21 Activity diagram for the Cu T i O system at 10000 C. . . . . . . . 132
5-22 Aluminum eect on the thermochemistry of zirconia/Cu T i Al
interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5-23 Cu 5 wt.% braze/zirconia interface, at 10500 C. . . . . . . . . . . . . 134
6-1
6-2
6-3
6-4
6-5
6-6
Calculated Cu N i and Cu F e phase diagrams. . . . . . . . . . .

Schematics of the simulation domain. . . . . . . . . . . . . . . . . . .
S/L interface position calculations. . . . . . . . . . . . . . . . . . . .
Diusion path calculations. . . . . . . . . . . . . . . . . . . . . . . .
Molar fraction prole for the diusion couple calculations. . . . . . .
Normalized chemical activities of T i at t = 0 and t = 100 s for the N i
and F e cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-7 Phase fraction calculation for Cu N i T i case, after t = 100 s and
T = 12000 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-8 ZrO2 /Cu T i/Inconel 718 r joints brazed at two dierent temperatures [51]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-9 X-Ray mapping of ZrO2 /Cu T i sample. . . . . . . . . . . . . . . .
12
142
145
146
147
149
150
151
151
153
7-1
7-2
7-3
7-4
7-5
7-6
7-7
7-8
7-9
7-10
7-11
7-12
7-13
7-14
7-15
7-16
7-17
7-18
7-19
7-20
7-21
7-22
(O) vs. xO for the Zr O and T i O systems. . . . . . . . . . . . 157

Calculated T i O phase diagram. . . . . . . . . . . . . . . . . . . . . 158
Stage I. of model by Torvund et al. [149]. . . . . . . . . . . . . . . . 159
Stage II. of model by Torvund et al. [149]. . . . . . . . . . . . . . . . 160
Free energy v.s composition diagram for the Ag Cu fcc system at
900 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
Gibbs free energy for the phases bcc T i(O), hcp T i(O), f cc T i(O)
(T iO1x ) at 12000 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
min GbccT i(O) , GhcpT i(O) , f cc T i(O) at 12000 C. . . . . . . . . . . 168
(O) vs. xO in T i O system. . . . . . . . . . . . . . . . . . . . . . . 168
Eect of k on the width of interfaces from Cahn-Hilliard equation. . . 170
Global mass error as a function of number of inner nodes. . . . . . . . 171
Eect of K on the global mass error of C-H numerical simulations of
coupled oxide growth. . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Concentration and chemical potential proles. . . . . . . . . . . . . . 173
Global mass conservation for the solution of the Cahn-Hilliard equation.174
Displacement of interfaces versus time and time1/2 . . . . . . . . . . . 175
Stefan condition at the hcp/bcc and f cc/hcp interfaces. . . . . . . . 177
Concentration proles at dierent times. . . . . . . . . . . . . . . . . 178
Chemical potential proles at dierent times for Cahn-Hilliard simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
2-D simulation of coupled oxide growth. Planar initial conditions. . . 179
2-D simulation of coupled oxide growth. Non-planar initial conditions. 180
SEM micrograph of a T i A55 sample oxidized at 6490 C for 65.9hr. 184
Calculated T i N phase diagram. . . . . . . . . . . . . . . . . . . . 185
Optical micrograph of the cross-sections of the ion-nitrided titanium
at 9000 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
A-1 Body-centered crystalline structure. . . . . . . . . . . . . . . . . . . . 200

D-1 Schematics of Metastable Equilibrium. . . . . . . . . . . . . . . . . . 224
D-2 Proposed diusion paths and experimental results by Paulasto et al.
[22]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
D-3 DGM R(CuT i2 ratio at the T i/( T i + Liquid 2) phase boundary as a
DGM R(CuT i)
function of T i content. . . . . . . . . . . . . . . . . . . . . . . . . . 228
E-1 Boundary conditions for simulations. . . . . . . . . . . . . . . . . . . 230
E-2 One-dimensional Newton method. . . . . . . . . . . . . . . . . . . . . 232
F-1 Experimental apparatus. . . . . . . . . . . . . . . . . . . . . . . . . . 235
F-2 Vacuum high-temperature furnace. . . . . . . . . . . . . . . . . . . . 236
13
14
List of Tables
2.1
2.2
2.3
Test of Eberhart model. Wetting of metals on sapphire (Al2 O3 ) . . .

Wettability and reactivity in Cu P d T i/oxide sessile drop experiments at 12000 C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contact angle and Wad in CuP d-T i/alumina wetting experiments. . .
3.1
3.2
3.3
Experimental and calculated invariant points in Ag T i system. . . . 46

Experimental and calculated invariant points in Ag Cu T i system. 53
Calculated and experimental invariant points in the CuT iZr system. 58
4.1
4.2
4.3
4.4
4.5
4.6
4.7
Calculated and experimental invariant points in the Zr O system. .

Experimental data for the Cu T i O system at 9450 C. . . . . . . .
Enthalpies and entropies of formation for the M6 X compounds. . . .
Structural information of T i4 Cu2 O and T i3 Cu3 O. . . . . . . . . . . .
Enthalpies and entropies of formation for the M6 X compounds. . . .
Parameters for dilute Cu O Zr solutions. . . . . . . . . . . . . .
Comparison between calculated and experimental invariant points in
the T i Zr O system. . . . . . . . . . . . . . . . . . . . . . . . . .
5.1
5.2
5.3
32
37
38
65
70
73
74
76
78
87
5.6
5.7
5.8
5.9
Ionic diusion coecients in Y SZ crystals. . . . . . . . . . . . . . . . 97

Diusion coecients in bcc and hcp T i Zr O solutions. . . . . . . 106
Calculated/estimated diusion coecients (cm2 /s) for phases in the
T i Zr O system at 17000 C. . . . . . . . . . . . . . . . . . . . . . 107
Titanium activities of Ag Cu T i Brazing Alloys. . . . . . . . . . 125
Normalized driving forces for precipitation at a metastable ZrO2 /Ag
Cu T i interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Compositions of points indicated in Fig. 5-16, at.% . . . . . . . . . 128
Thermochemistry of Cu Zr O melts at 11000 C in the Cu-rich region.136
6.1
Dissolution enthalpies of T i in melts. . . . . . . . . . . . . . . . . . . 141
C.1
C.2
C.3
C.4
C.5
Thermodynamic parameters for the ionic liquid phase.

Thermodynamic parameters for the bcc phase. . . . . .
Thermodynamic parameters for the f cc phase. . . . . .
Thermodynamic parameters for the hcp phase. . . . . .
Body-centered crystalline structure. . . . . . . . . . . .
5.4
5.5
15
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
209
211
212
213
214
C.6 Thermodynamic parameters for the Cu4 T i phase. . . . . . . . . . . . 215

C.7 Thermodynamic parameters for the CuM2 phase. . . . . . . . . . . . 216
C.8 Thermodynamic parameters for stoichiometric intermetallic compounds.216
C.9 Thermodynamic parameters for the T iO phase. . . . . . . . . . . . . 217
C.10 Thermodynamic parameters for the T iO2 (rutile) phase. . . . . . . . 217
C.11 Thermodynamic parameters for the ZrO2 phase. . . . . . . . . . . . 217
C.12 Thermodynamic parameters for the ZrO2 phase [52]. . . . . . . . . 218
C.13 Thermodynamic parameters for the ZrO2 (rutile) phase [52]. . . . . 218
C.14 Thermodynamic parameters for the 3 3 M6 X compound. . . . . . . 219
C.15 Thermodynamic parameters for stoichiometric oxides. . . . . . . . . . 219
C.16 Functions used in the models. . . . . . . . . . . . . . . . . . . . . . . 220
16
Nomenclature
id
G Ideal mixing contribution to the total Gibbs energy of phase , . Units:

m
see equation (A.4), page 177
xs
G Excess contribution to the total Gibbs energy of phase . Units:

m
equation (A.4), page 177
J
mol
J
,
mol
, see
(A, B)i (C, D)j Sublattice notation. Phase with two sublattices with i and j sites in
each sublattice. Elements A, B in rst sublattice, while C, D are in second
sublattice. , see equation (3.0), page 24
A
Solute A dissolved in solvent , see equation (2.5), page 7
aA
Chemical activity of element A , see equation (2.6), page 7
Aj
Partial quantity from molar quantity Am , see equation (A.21), page 183
Cj
Molar concentration. Units:
n
Dkj
Diusion coecient in the (n 1) by (n 1) diusion matrix. Units:

equation (6.2), page 123
Electron.
ej
i
Interaction coecient of j over i in solvent M , on a weight percent basis , see

equation (4.8), page 57
Eg
Bandgap Energy, eV , see equation (2.15), page 13
Free energy of system , see equation (7.4), page 142
Gibbs free energy of phase , see equation (A.1)
G
ij
Gibbs free energy of compound ij in crystalline structure . Units:

equation (A.3), page 177
mol
m3
, see equation (6.2), page 123

m2
s
J
mol
, see
, see
GHSERi Gibbs energy with respect to enthalpy of standard element reference state
J
(SER). Units mol , see equation (A.3), page 177
h
Computational grid spacing.
Ii
First ionization potential , see equation (2.14), page 11

17
mol
m2 s
Jk
Mass ux. Units:
Gradient energy coecient , see equation (7.10), page 144
KD (T ) Equilibrium constant for oxide dissolution reaction , see equation (2.6), page 7
Keq
Equilibrium constant , see equation (3.2), page 28
L
A:C,D Interaction parameter between C and D in second sublattice of phase when
A is present in second sublattice ., see equation (A.17), page 182
L
i,j,k The ternary interaction parameter between elements i, j, k in substitutional
phase , see equation (3.4), page 34
Lij
Regular solution interaction parameter between components i and j. Units:

J
, see equation (A.6), page 178
mol
Mi
Molecular weight of element i.
nh
Hole concentration in valence band , see equation (2.15), page 13
n0
h
Constant.
X
OO
Oxygen in oxygen site , see equation (5.0), page 77
PI0 (Y ) Zeroth-order product of site fractions ., see equation (A.15), page 181
pO2
Oxygen partial pressure , see equation (3.2), page 28
J
Universal gas constant. R = 8.314 mol K
Interatomic distance at ceramic/metal interface , see equation (2.14), page 11
Temperature, K
Time.
Velocity of interface . Units: m/s , see equation (6.3), page 123
VO
Oxygen vacancy , see equation (5.0), page 77
Va
Vacancy.
Wad
Work of adhesion. Units: J/m2 , see equation (2.2), page 6
xi
Mole fraction of component i.
xD
O
Molar fraction of oxygen dissolved in the melt when oxide Am On is in equilibrium with it , see equation (2.9), page 8
xP
O
Molar fraction of oxygen in melt when it is at equilibrium with reaction product

18
n
yi 1
0
Site fraction of element i in sublattice n ., see equation (A.12), page 181
G
(i)
Gibbs free energy of the reference compound ij in the crystal structure .,

see equation (A.14), page 181
1 (j)1
Polarizability of element i , see equation (2.14), page 11
0 G Gibbs free energy of formation for phase

f
GInt Gibbs free energy for interface formation , see equation (2.1), page 6
f
A
0 Gf m On Gibbs free energy of formation of Am On oxide. Units: J/mol , see equation (2.6), page 7
M
0 Hf O Enthalpy of formation of M O oxide , see equation (2.14), page 11
j
i
Interface thickness , see equation (7.18), page 150

Wagner rst order interaction coecient between element i and element j
dissolved in solvent M .
Chemical activity coecient for element i.
Chemical activity coecient for element i at innite dilution.
Hij,f Standard enthalpy of formation of compound ij. Units:

tion (A.3), page 177
(i)
, see equa-
J
molK
, see equa-
Chemical potential of element i.
J
mol
Molar volume , see equation (7.4), page 142
Sij,f Standard entropy of formation of compound ij. Units:

tion (A.3), page 177
J
m2
Surface energy. Units:
LV
Surface energy of liquid/vapor interface. Units: J/m2 , see equation (2.1),

page 6
SL
Solid/liquid interfacial energy. Units: J/m2 , see equation (2.1), page 6
SV
Surface energy of solid/vapor interface. Units: J/m2 , see equation (2.1),

page 6
Contact angle, in degrees , see equation (2.3), page 6
KP (T ) Equilibrium constant for precipitation , see equation (2.12), page 10
19
20
Chapter 1
Introduction
A wide range of ceramic/metal interfaces occurs naturally. In general, thin oxide
layers cover the surfaces of most metals under normal conditions. This metal/oxide
system constitutes an excellent example of a typical ceramic/metal C/M interface.
This work, however, is concerned with synthetic interfaces created during ceramic/metal
joining, coating processes and so on. Understanding of the thermo-kinetic phenomena
occurring at ceramic metal interfaces comprises three dierent levels:
i Thermodynamic interactions between the dierent phases present in the system.

This implies the determination of the driving forces for any further interfacial
interaction, such as dissolution, diusion, precipitation of reaction products, etc.
ii Once the thermodynamic driving forces have been determined, it is possible to
analyze the kinetic processes occurring at C/M interfaces, such as charge and/or
mass transfer.
iii Determination/prediction of the microstructural evolution of the interfaces when
subjected to arbitrary thermo-mechanical and chemical conditions.
These three hierarchical levels of analysis warrant varying degrees of complexity, depending on the number of elements and phases involved, stoichiometry of the
ceramic and possible reaction products, nature of transport mechanisms across interfaces, moving boundaries, morphological instabilities, and so forth.
In this work, a methodology for the study of the thermochemical interactions at
ceramic/metal interfaces is proposed. In order to illustrate the concepts developed in
this thesis, the nature of the interfacial interactions between zirconia-based ceramics
and active metals was selected. The selection of zirconia/active metal interfaces was
based on the importance of these ceramic/metal systems in established and emerging
technologies, such as solid oxide fuel cells, thermal barrier coatings for turbine blades,
biomedical materials, structural ceramic/metal composites, etc.
21
Through the use of thermodynamic models based on the CALPHAD formalism it

will be shown that it is possible to perform metastable thermodynamic calculations
that allow the prediction of diusion paths across zirconia/active-metal interfaces.
Additionally, equilibrium calculations of activity diagrams will be applied to the understanding of the complex interfacial reactions occurring during active metal brazing of zirconia-based ceramics. By using simple inter-diusion simulations, it will be
demonstrated that the base metal in ceramic/metal joints may play an essential role
in determining the thermochemical interactions at the ceramic/metal interface. In
many cases, the morphology of the reaction layers formed at ceramic/metal interfaces
determines the nal properties. Thermodynamic calculations and one-dimensional
diusion simulations are incapable of describing the morphological evolution of these
interfaces. In this thesis, it will be demonstrated that thermodynamic models of the
likely reaction products at zirconia/metal interfaces can be coupled to kinetic models
using the diuse-interface formalism to describe the formation and evolution of complex ceramic/metal interfacial structures.
The organization of this work is as follows:
In Chapter 2, the main physico-chemical aspects of C/M interfaces are discussed.
A review of the most relevant work on the nature and characteristics of C/M interfaces is presented. Subsequently, the implications of these theories on the problem at
hand: the creation of synthetic C/M interfaces is discussed.
Although ceramic/metal interfaces have been experimentally investigated for several decades, thermodynamic models to describe even the most simple systems are
still lacking. For the problem at hand, it was found that there were no thermodynamic assessments for many of the systems that were needed to achieve a complete
thermodynamic description of the system under study.
Despite the widespread use of Ag Cu T i alloys in ceramic/brazing applications, no thermodynamic assessment for this system has been published before. In
Chapter 3, the thermodynamic model for this system is presented and the available
experimental information is critically assessed. In addition to the Ag Cu T i system, active brazes based on amorphous Cu T i Zr alloys are becoming important
in ceramic/metal brazing applications, mainly due to the higher operating temperature of this system. Chapter 3 presents a thermodynamic assessment for this system.
Despite the fact that it has been used commercially for several decades, there
has been very limited thermodynamic investigations of the Zr O system, from
the CALPHAD point of view. In Chapter 4 a thermodynamic assessment for this
system is presented. Additionally, a previously developed thermodynamic model for
the Cu T i O system is rened. In Chapter 4, a thermodynamic description for
the T i Zr O system is also developed. These last systems are of fundamental importance for the understanding of zirconia/active-metal interactions. Finally, a
critical assessment of the scarce thermochemical data for the CuOZr is presented.
22
In Chapter 5, the thermodynamic models developed in Chapter 3 and Chapter 4

are used to describe the complex reaction sequences observed in experiments on ceramic/metal interactions. In the case of ZrO2 /T i interactions, metastable thermodynamic calculations are used to determine the likely diusion paths in both solid/solid
and solid/liquid interfacial processes. Activity diagrams calculated based on the
model for the Cu T i O system are used to understand the interfacial structure
resulting from active brazing experiments of zirconia to metals and to itself.
In Chapter 6, the eect of the substrate metal on the thermochemistry of the active
brazing alloy during ceramic/metal brazing operations is studied. By using simple
diusion simulations, the interdiusion processes occurring at the metal/braze interface are elucidated and quantied.
In Chapter 7, the coupled growth of typical reaction layers formed due to ZrO2 /T i
solid state diusional reactions are investigated using the Phase-Field approach. By
combining semi-empirical thermodynamic models with the diuse interface formalism, it is shown that the reaction layer sequencing and morphology observed in the
zirconia/titanium system can be explained, at least in a qualitative manner. In this
chapter is also shown that the model developed can be generalized to other complex
interfacial phenomena in other systems.
23
24
Chapter 2
Ceramic/Metal Interfaces: Basic
Ideas, Joining Processes
2.1
Introduction
Ceramics and metals have dierent electronic structures [1]. While the bonding electrons in metals are highly delocalized, electrons in ceramics are highly localized around
the ions in the crystalline lattice. Since the nature of the inter-atomic bonding, as
well as the electronic density, change abruptly at C/M interfaces, it is to be expected
that these interfaces will be energetically unstable .
In this chapter, the main physico-chemical aspects of C/M interfaces are discussed. A review of the most relevant works on the nature and characteristics of
C/M interfaces will be presented. Subsequently, the implications of these theories on
the problem at hand: the creation of synthetic C/M interfaces will be discussed.
2.2
Wetting as a Metric of Ceramic/Metal Interfacial Energetics
Quantication of the energetics of ceramic/metal interfacial interactions is only possible when the proper metric is used. In the solid/solid case, a measure of the strength
of such interactions could be the shear stress necessary to separate the C/M interface. This metric, however, is of little practical use in this case because the mechanical
strength of a ceramic/metal interface is not only determined by interfacial energetics,
or adhesion, but also by the mechanical state (i.e. stored elastic/plastic energy due
to dierences in thermo-mechanical properties) of the bulk materials themselves (see
for example Park et al. [2]). Since the nal state of the system is a function of the
actual process of C/M interface formation, it is dicult to de-couple the energetics
of C/M interfacial adhesion from the mechanical state of the system.
This statement refers to non-reactive metal/ceramic systems. In the case of reactive systems,
the chemical anity between the ceramic cations and the active metal is such that a reaction product
forms at the C/M interface. It is highly likely that this reaction product in turn has intermediate
properties to those of the reactive alloy and the ceramic.
The metal could be either in the solid or the liquid state
25
This problem becomes more tractable in the case of solid/liquid interfaces, since
the metric of the interfacial energetics is much easier to access and is not coupled to
other (mechanical) energetic contributions to the total energy of the system. This
metric is the wetting behavior of the solid/liquid system.
In general, the driving force for the formation of a ceramic/metal interface is the
decrease in free energy that occurs when intimate contact is established between the
two substrates [3]. This change in energy (per unit area) is given by:
GInt = SV + LV SL
f
(2.1)
where GInt is the Gibbs free energy for interface formation, and SV and LV are
f
the surface energies of the solid and the liquid, while SL represents the interfacial
energy of the Solid/Liquid interface. In the case when only chemical bonding has occurred and the ceramic and metal remain free of plastic deformation, G is identical
to the work of adhesion, Wad , necessary to separate a unit area of interface into the
two original surfaces [4]:
Wad = SV + LV SL
(2.2)
As indicated by Eq. 2.2, (known as the Dupr equation) the work of adhesion,
e
Wad , increases as the S/L interfacial energy becomes smaller than the surface energies
of the solid and liquid phases. A greater Wad implies a greater S/L interfacial stability.
If the system is considered to be at equilibrium, the surface and interfacial energies balance each other and the following relationship can be established:
SV = Sl + LV cos
(2.3)
Eq. 2.3 is also known as the Youngs equation , and it can be combined with
Eq. 2.2 so the work of adhesion, Wad , can be related to the more accessible contact
angle, , between the liquid metal and the solid substrate:
Wad = LV (1 + cos )
(2.4)
The de-coupling of adhesion from the total strain energy of a liquid/solid C/M interface occurs
because the liquid cannot be subjected to shear stresses.
Note that Youngs equation is only an approximation and should be used with care when
considering the real 3D-shape of liquid droplets wetting a surface. Moreover, Youngs equation only
applies at equilibrium conditions.
26
By convention, wetting can be dened as the condition at which the contact angle
between the liquid metal and the solid substrate is below 900 . As 00 , complete
spreading occurs, and the work of adhesion reaches the maximum allowable value,
namely 2 LV .
By measuring the contact angle between a liquid and a solid substrate, it is thus
possible to establish the stability of the C/M interface. In the following section, the
wettability of liquid metals over ceramic substrates will be addressed from the point
of view of the the reactivity or non-reactivity of the metal.
2.3
Reactive vs. Non-Reactive Wetting
At high temperatures, there is always a certain degree of dissolution of a ceramic substrate when is in contact with a liquid metal, regardless of the reactivity of the latter.
To establish a relationship between the reactivity (or chemical activity) of the liquid
metal and the thermodynamic driving force for the formation of a ceramic/metal
interface, one has to establish rst a quantitative measure of the reactivity of the
system. The extent of oxide dissolution in the liquid metal provides with a good
indication of the latters reactivity [5, pp. 198-199]. In this section, the dissolution
of a ceramic substrate in a liquid metal will be considered under both reactive and
non-reactive conditions, and a relationship between reactivity and wettability will be
presented.
2.3.1
Ceramic Dissolution: Non-reactive Case
Consider for example the following dissolution reaction:

1
n
Am On A + O
m
m
where Am On is the ceramic substrate and M is the liquid metal.
(2.5)
The equilibrium constant, KD (T ) for this reaction [5, p. 199 ] is given by:
n/
KD (T ) = aA aO m = exp
+0 GAm On
f
m RT
(2.6)
Assuming constant chemical activity coecients, i i
aA = xA A aO = xO O
(2.7)
Eq. 2.6 can then be written as follows:

n
KD (T ) = xA A {xO O } /m
(2.8)
If the only source of elements A and O is the oxide substrate, the molar fractions
27
of O and A are then related by m xO = n xA .

The equilibrium molar fraction of oxygen, xD is given by:
O
xD =
O
n
m
m
m+n
where
KD (T ) =
KD (T ) m+n
(2.9)
KD (T )
n
A (O ) /m
Eustathopoulos et al. [5] have compiled a large number of ceramic oxide/metal

wetting experiments in which the contact angle between the liquid and the solid
substrate has been measured using the sessile drop technique. By calculating the
oxygen molar fraction in the liquid metal (xD ) using homogeneous thermodynamic
O
properties for the oxides and approximate values for the activity coecients of both
oxygen and the cation of the ceramic in the melts, Eustathopoulos et al. [5] found
a relationship between the extent of oxide dissolution, as measured by xD , and the
O
contact angle at the ceramic/metal junction. Their ndings can be summarized as
follows (see Fig. 2-1):
130 0
Contact Angle
90 0
50 0
Non Wetting
Wetting
00
10 -5
10 -4
10 -2
Oxygen Dissolution in the Liquid Metal
Figure 2-1: Schematic of experimental compilation by Eustathopoulos et al. [5]. This
diagram is not at scale.
i) For the range xD < 105 , the observed contact angle was virtually independent
O
of the extent of dissolution of the ceramic oxide in the metal. The contact angles
observed under these conditions were within the 1000 < < 1400 range.
ii) As the extent of oxide dissolution increased above the xD > 105 value, the
O
contact angle started to decrease.
28
iii) At about xD 104 , all the sessile drop experiments that were examined exO
hibited a contact angle of 900 , which corresponds to the wetting/non-wetting
transition.
iv) At even greater oxide dissolution, the contact angle approaches 00 degrees. These
low contact angles, however, were due to the formation of a wettable reaction
layer between the liquid metal and the ceramic substrate. The liquid metals in
these cases were invariably highly reactive and the free energy of formation of
their oxides was highly negative.
From the observations by Eustathopoulos et al. [5], the limit between wetting
and non-wetting seems to be xD 105 . It is important to note that the calculated
O
extents of dissolution did not consider the eect that the furnace atmosphere has on
the chemistry of the liquid metal. For extremely reactive systems, providing there
is a low enough pO2 surrounding the samples, the eect of the furnace atmosphere
is not expected to aect the thermochemistry at the ceramic/metal interface. For
non-reactive systems,however, the eect can be signicant [5].
2.3.2
Addition of Reactive Elements
In general, even if it is possible to reduce the contact angle observed in non-reactive

metal systems by the addition of solutes that decrease the surface tension of the
liquid [5, section 6.5] or promote the segregation of oxygen to the ceramic/metal
interface (see Section 2.4), the contact angle between a liquid metal and a ceramic
substrate cannot be reduced to values below 600 unless active elements are added
to the liquid.
One can consider again the dissolution reaction [5]:
1
n
Am On (A)M + (O)
(2.10)
m
m
If a second element, Re is added to the liquid M and this element has a signicant
oxygen anity, Eq. 2.9 can be modied in the following manner:
xD
O
n
=
m
m
m+n
KD (T ) m+n exp
n
Re xRe
m+n O
(2.11)
where Re is the Wagners rst-order interaction coecient between O and Re in liqO

uid M .
According to Eq. 2.11, if Re < 0, the amount of oxygen in the liquid M Re
O
that is in equilibrium with the oxide Am On increases rapidly as the amount of B increases. If the interaction between the reactive element R and oxygen is very negative
Re << 0, the precipitation of an oxide Rey Oz can be promoted:
O
It has been assumed that the remaining interaction coecients are negligible compared to Re .
O
29
yRe + zO Rey Oz
The equilibrium constant for this reaction, KP (T ) will then be given by:
Re Oz
KP (T ) = {xRe Re }
{xO O }z
0 Gf y
RT
= exp
(2.12)
The mole fraction of oxygen in the liquid M in equilibrium with the reaction
product Rey Oz is given by:
y/z
xP = KP (T ) xRe exp O xRe

O
Zr
(2.13)
where
y/z
KP (T ) = KP (T )1/z Re (O )1
15. 5
16. 6
16. 8
16
17
17. 2
Re/Re yO z
17. 4
log(x(O))
log(x(O))
16. 5
17
Re/Re yO z
17. 6
17. 8
17. 5
18
18. 2
A/Am O n
18
A/Am O n
18. 4
18. 5
5
4. 5
3. 5
log(x(Re))
2. 5
1. 5
(a) Slightly reactive element addition.

Graphical representation of Eq. 2.11 and
Eq. 2.13.
18. 6
5
4. 5
3. 5
log(x(Re))
2. 5
1. 5
(b) Strongly reactive element addition.

Graphical representation of Eq. 2.11 and
Eq. 2.13.
Figure 2-2: Eects of active element additions on the dissolution of ceramic oxide substrates.
Fig. 2-2 is a graphical representation of the two dierent behaviors that can be
expected when active elements are added to non-reactive liquid metals in contact
with ceramic substrates. In both cases, it can be seen that the ceramic dissolution increases dramatically as the concentration of the active element in the liquid increases.
In Fig. 2-2(a) the active element, however, does not react with the oxygen dissolved
in the metal to form a reaction product, as the A/Am On and Re/Rey Oz equilibrium
curves never intersect [5, p. 249]. Fig. 2-2(b) illustrates the case in which the reactive element has such a strong anity for oxygen (Re << 0), that the precipitation
O
of a Rey Oz oxide occurs, as is illustrated by the interception of the A/Am On and
Re/Rey Oz equilibrium curves. It is to be expected that, while the active element
30
addition in Fig. 2-2(a) will decrease the contact angle in a noticeable manner, the
precipitation of a reaction product, as shown in Fig. 2-2(b), will promote the greatest
wettability, provided the reaction product is wettable.
In the following section, a brief review of the most successful description of the energetics of ceramic/metal interfaces will be presented. Wettability will then be related
to the intrinsic electronic structure of ceramics and metals. Although the models are
principally focused on the interaction of non-reactive liquids on ceramic substrates,
it will be seen (see Section 2.6) that the models can also explain the dierent degrees
of wettability observed in reactive C/M systems.
2.4
Non-reactive Metal-Ceramic Wetting: Inuence of the Electronic

Structure
Although there have been several attempts at understanding the nature of ceramic/metal
interactions, there are very few models capable of describing the full set of trends and
behaviors that have been observed in experiments. For example, early attempts to
describe the metal M - oxide AO interaction at the C/M interface focused on nding a correlation between the energetics of the interface (Wad , for example) and the
formation enthalpy or Gibbs energy of formation of the M O oxide [5]. Such a correlation would indicate a chemical interaction between the ceramic substrate AO and
the metal M at the interface, even though a chemical reaction would not necessarily
occur. According to these models, Wad (see Section 2.2) would be related to the
bond energies between the anion and the metal. One such model was proposed by
MacDonald and Eberhart [6]:
M
Wad = a0 Hf O + b
M O
IM IO
r6
IM + IO
(2.14)
0
where Hf,M O is the enthalpy of formation of the oxide M O, i are the polarizabilities of the oxygen and metal atoms, Ii are the rst ionization potentials, r is the
interatomic distance at the ceramic/metal interface and a and b are proportionality
constants.
M
This model requires the addition of a chemical (a0 Hf O ) and a physical
IO
(b M6 O IIM+IO ) interaction term. The chemical interaction term is the enthalpy
r
M
of formation of the liquid metal oxide. The physical interaction term takes into account a dipole-dipole interaction between the ceramic anion and the liquid metal and
is essentially constant for most metals [7].
Table 2.1 shows that the McDonald-Eberhart model does not predict the correct
tendencies according to Eq. 2.14. Note that while the enthalpy of formation of SnO is
twice that of CuO, the work of adhesion at the Sn/Al2 O3 is half that of the Cu/Al2 O3
case. It is evident that this model (and all models that try to correlate M O bond
31
Table 2.1: Test of Eberhart model. Wetting of metals on sapphire (Al2 O3 )
Metal
Contact
Angle [7]
Work of
adhesion mJ [7]
m2
Cu
Sn
130
122
Formation
enthalpy [8]
kJ/mol
156
287
450
225
energies to Wad ) does not agree with the observed experimental facts.
One of the best models for describing non-reactive ceramic/metal C/M interfacial
interactions was proposed by Li [9]. He examined the dependency of the work of adhesion, Wad , of a metal/oxide system on the electronic properties of both the (liquid)
metal and the ceramic substrate. In his analysis, Li [9] excluded metallic systems
that reacted with ceramic substrates forming reaction products, since this necessarily
changes the nature of the original C/M interface. From this analysis, a correlation
between the electronic properties of the C/M system and the work of adhesion was
found (See Fig. 2-3).
1500
3000
Al2O3
1000
500
Al
Cu
500
Pd
Au
Ga
In
Pb
Mn
Si
Wad mJ/m
MgO
BeO
CaO
TiO2
ZrO2
UO2
Cr2O3
Wad mJ/m
1500
0
0
Co
1000
2000
Fe
Ni
NiO
CoO
2500
Pt
Ni
Fe
Co
0
0
Sn
1
nws in d. u. 1 d. u. = 6.0x10
Bang Gap Energy for Oxide Eg, eV
(a) Work of adhesion Wad of molten Co,

F e, N i as a function of the bandgap energy Eg of dierent oxides [9]. The data
was originally presented as Wad vs. G0 .
f
It was modied using the plots of Eg vs.
G0 found also in Li [9]
f
22
el/cm 3
(b) Work of adhesion Wad of dierent liquid metals with alumina vs. the electron density nws at the boundary of the
Wigner-Seitz cell.
Figure 2-3: Correlation between the work of adhesion Wad of a ceramic/metal interface and
the electronic properties of both ceramic substrates and liquid metals.
Based on the correlations depicted in Fig. 2-3, Li [9] formulated a model to

describe the energetics of C/M interactions between non-reactive metals and ioniccovalent oxides as a function of the electronic properties of both bulk materials.
32
According to this model, the stability of a non-reactive C/M interface is increased

by greater electron transfer from the metal conduction band into the oxide valence
band. When a metal comes into contact with a ceramic surface (in the case examined
by Li [9], an oxide), the thermally created holes in the valence band can be lled by
the free electrons of the conduction band of the metallic lattice. The ceramic surface
can thus be stabilized by the presence of a foreign metallic phase. This interaction is
metallic-like in nature and therefore is expected to have long range character, as has
been observed in some rst-principle calculations [5, chap. 6]. Such a transfer would
be favored by a large electronic density, nws on the metal side of the C/M interface .
Fig. 2-3(b) shows how the work of adhesion between several molten non-reactive metals and alumina changes with nws . As can be seen, Lis model agrees with the trends
observed in Fig. 2-3(b).
Another implication of the model by Li [9] is that electron transfer across a C/M
interface will be favored in cases for which the valence band in the ceramic material
has a large concentration of holes. For an idealized oxide crystal, free of impurities,
the concentration of holes in the valence band is given by:
Eg
nh = n0 exp 2 Kb T
h
(2.15)
where Eg is the bandgap energy. According to Eq. 2.15, as Eg increases, the concentration of holes nh decreases at an exponential rate. According to Lis model,
insulating materials with large Eg would then interact in a weakly manner with liquid metals. This behavior can be observed in Fig. 2-3(a). Another conclusion from
this model is that Wad ( would decrease) would increase with temperature, as was
experimentally observed by Rivollet [11].
In general, Eg is correlated to the Gibbs free energy of formation G0 of the
f
ceramic (See Fig. 2-4). Thus, a correlation between ceramic stability and work of
adhesion should then be observed. According to the experimental data presented
by Li [9], the work of adhesion of the corresponding C/M interface decreases as the
Gibbs free energy of formation of the oxide increases (using the same liquid metal).
Although the model described above focused on the interactions between nonreactive liquids and metals, it is possible to extend the conclusions of the model to
solid/solid systems. In fact, Li [9] found experimental evidence [14] that showed a
linear correlation between the work of adhesion of solid metal/Al2 O3 interfaces, and
the electronic density of the respective metals, nws (see Fig. 2-3(b)). This means
that the nature of bonding at solid/solid C/M interfaces is very similar to that in
liquid/solid C/M systems. Despite the similarities, it is important to consider that,
The electronic density, nws , is one of the terms used in Miedemas model to describe cohesion
in metallic systems and it corresponds to the electron density at boundaries of the Wigner-Seitz
cells [10].
33
Eg=0.0121Gf+0.1095
CaO
Al2O3
B2O3
Eg, eV
MgO
MoO3
TiO2
2
1
0
0
NiO
PdO
PtO2
100
200
300
400
500
600
700
Gibbs Free Energy of Formation Gf (kJ/g-atom oxygen).
Figure 2-4: Relationship between bandgap energies [12] and Gibbs free energy of formation [13] for several oxide ceramics
contrary to S/L interfaces, in the case of S/S interfaces large strain energies can be
generated at the ceramic/metal interface as it is cooled from the joining temperature.
These strain energies arise due to dierences in the thermo-mechanical properties of
the materials. The addition of this extra strain energy can result in an actual rupture
of the interface if its value is greater than the interfacial work of adhesion.
It is important to note that the model described above only applies when considering pure metal-ceramic interactions, with no surface contamination and under
high-vacuum conditions. Adsorption of surface active elements, such as oxygen, onto
the liquid metal surface will decrease the liquid surface tension and, according to
Youngs equation (Eq. 2.3), decrease . Additionally, alloying the liquid metal with
another metallic solute with much smaller surface tension and a strong tendency for
segregation will also decrease the contact angle between the molten metal and the
ceramic.
The model described in this section applies only to non-reactive C/M interactions
for the simple reason that the product of a reactive system would permanently alter the nature of the original C/M interface. With an interfacial reaction product,
a new set of interfaces would have to be considered and the analysis would then be
more complicated than the simple model outlined above. Nevertheless, we can expect
that the resulting C/M interface can be analyzed under the assumptions of Lis model.
2.5
Joining Ceramics to Metals
Joining facilitates the use of structural ceramics -as well as other types of ceramic
materials- by providing a means to manufacture structures which cannot be made in
one piece or which can be made less expensively by joining. Because of their intrinsic
34
brittleness, most ceramics are fabricated using diverse joining techniques, thus avoiding large mechanical loads involved in bulk processing.
Furthermore, ceramic/metal, C/M, joining is particularly useful when trying to
combine the mechanical and/or chemical properties of both ceramics and metals in
the same device -i.e. C/M composites, protective coatings, etc. Of all the current
C/M joining applications, one of the most important is the attachment of ceramic
components operating at high temperatures to structures or moving parts which must
withstand stresses or temperature gradients too great for ceramics, and which are usually made of metals.
When designing a C/M joint, one always has to face the fact that the dierent
nature of the materials makes the joining process a dicult task. In order to optimize
the design of a successful C/M joint it is crucial to overcome two dierent although
inter-related problems:
C/M interfaces tend to be thermodynamically unstable. Since the electronic
structures of the ceramic and metallic lattices are dierent (see Section 2.4), the
ceramic/metal interfacial energy is usually greater than the free surface energy
of the ceramic itself. Therefore, the ceramic free surface is usually more stable
than a C/M interface.
C/M interfaces are mechanically unstable . Because of the usually dierent
mechanical and thermo-mechanical properties of the ceramics and metals, C/M
interfaces are usually subjected to large stresses, especially when the joint must
operate at high temperatures.
The resulting C/M interfaces usually have to be stable over a wide range of temperature conditions. At the same time, the joint must be able to withstand the large
stresses generated due to the dierent thermo-mechanical properties of both ceramics
and metals.
As a way of summarizing the broad spectrum of possible C/M joining techniques,
one can classify them into one of the following major groups:
i) Mechanical joining: Used in both traditional and new applications such as securing of furnace roof refractories with metal hooks. Press-and-shrink tting is
another type of mechanical joining technique that is widely employed in mass
production processes.
ii) Direct joining: Achieved by pressing together very at surfaces to achieve bonding via solid-state diusion processes. In these type of processes, no intermediate
materials are needed to create a joint between dissimilar materials.
This is only true when both substrates are in the solid steta
35
iii) Indirect joining: This is the most common method for obtaining high-integrity
joints. It uses a wide range of intermediate materials such as organic adhesives,
glasses or glass-ceramic composites, oxide mixtures, or metals. Metal interlayers
are used in both solid and liquid phase diusion processes, brazing, etc. In these
techniques, pretreatment of the ceramic surface is sometimes needed to guarantee
good wetting properties.
The choice of a particular joining technique, as has been noted above, depends on
the nature of the ceramic and the metal, as well as on the operating conditions for
which the joint is being designed. Nevertheless, no matter which process is chosen,
the formation of a successful joint depends on the achievement of intimate contact
between the work pieces; the conversion of these contacting surfaces into an atomically bonded interface; and nally, on the ability of this interface to accommodate
thermally-induced stresses which are generated either during cooling from the joining
temperature or by temperature gradients due to the normal operating conditions of
the part.
One of the purposes of this thesis is to investigate some of the basic aspects of the
thermochemical interactions occurring at C/M interfaces. It is hoped that this understanding will make a contribution to the development of synthetic C/M interfaces
of practical use. Due to the complexity of this problem, it is necessary to narrow the
scope of this work, while at the same time extracting general principles that would
allow a better design and understanding of practical C/M interfaces. C/M active
metal brazing (to be described in Section 2.6) has been chosen as the primary example. Specically, the thermochemical phenomena occurring at C/M brazed joints will
be addressed. By choosing relevant C/M systems, it is expected that generalizations
applicable not only to C/M joining but to the more general eld of C/M interfacial thermochemistry can be made. While C/m active brazing is the main example
studied in this work, other instances where ceramic/metal interactions are important
are examined. In all the cases explored zirconia has been selected as the ceramic
substrate, for reasons outlined elsewhere in this thesis (see Section 5.2).
2.6
Active Brazing
According to Lis model (See Section 2.4) the electronic structure of both the ceramic
and the metal substrates determine the energetics of ceramic/metal C/M interfaces.
In general, non reactive-metal/ceramic systems show contact angles above 900 . This
large contact angle implies that most non-reactive C/M interfaces would tend to be
energetically unstable. If one were to create a synthetic C/M interface by indirect
joining of a ceramic substrate to a metal part (as outlined in Section 2.5) it would
be necessary to change the energetics of the resulting interfacial interactions in such
a way that a stable interface can be created.
Of the ceramic/metal joining techniques described in Section 2.5, one of the most
widely used is active brazing. In general, brazing can be dened as any joining tech36
nique that relies on the presence of a liquid interlayer between two solid substrates.
The interlayer should have a melting point lower than those of the joined substrates.
After the interlayer solidies, a joint is created, whose mechanical properties will be
determined by the adhesion between the interlayer and the substrates as well as by
the resulting mechanical stresses generated when the interface is cooled from the joining temperature .
As noted above, common brazing techniques cannot be applied to C/M joining
applications because the metals normally used in these techniques (copper, silver,
nickel) do not wet most ceramic substrates . Fortunately, these alloying systems can
be modied to promote wetting ( << 900 ) on even the most stable ionically bonded
ceramic oxides (such as Al2 O3 , ZrO2 , etc.) [15].
One of theories that have been developed to explain the benecial eects of adding
active elements to non-reactive brazing alloys is that the evolved heat from the dissolution and precipitation reactions constitutes a negative contribution to the formation
energy of the C/M interface. If this were the only contribution of reactive brazing,
a greater C/M work of adhesion (or smaller contact angle ) would result as the
interfacial interaction became more intense. In general the intensity of the C/M
interfacial interaction could be indicated by the relative thickness of the interfacial
reaction product. Following this reasoning, a greater reaction layer thickness would
have to be associated with a smaller contact angle . Table 2.2 shows that, at least
for some systems, this is not the case.
Table 2.2: Wettability and interfacial chemistry of Cu P d T i/oxide sessile drop experiment at T = 12000 C [16].
Substrate
Alumina
Mullite
Silica
Contact
Angle
34
32
35
Interfacial Thickness,
Product
m
T i2 O 3
0.5
T i2 O 3
1.0
T i2 O 3
10.0
Although in many cases the presence of a continuous reaction layer promotes wetting in C/M systems, there are some cases in which this is not the case. One example
of these situations is the sessile droplet experiments of liquid Ag over a SiC substrate
reported by [17]. At high temperatures (T 11000 C), Ag dissolves SiC forming Si
dissolved in Ag and graphite. Before graphite is formed, the molten Ag wets the SiC
substrate ( << 900 ). Nevertheless, as soon as a continuous graphite layer forms at
These mechanical stresses originate from the dierent thermo-mechanical properties of the
materials to be joined, such as coecient of thermal expansion, elastic moduli, etc.
This non-wetting is particularly evident for the most thermodynamically-stable ceramics, such
as highly ionic oxides (See Section 2.4).
37
the ceramic/braze interface, de-wetting occurs ( >> 900 ).
Table 2.3: Contact angle and Wad of CuP d-T i/alumina sessile drop experiments [18]
Wad ,
mJ
m2
T i content
in braze
Oxide
Bonding
Character
Contact
Angle
Al2 O3
Insulating
127
560
T i3 O5
Semi-metallic
92
1350
T i2 O3
Semi-metallic
70
1880
24.7
T iO1.2
Metallic
18
2730
A greater insight on the benecial eects of active element additions to brazing

alloys can be gained by examining Table 2.3. From this table, it is apparent that
the nature of the reaction layer (wether is conducting or insulating) formed due to
the interaction between the CuP d-T i braze alloy and the alumina substrate has a
great inuence on the C/M interfacial adhesion. This result can be examined in
light of the model developed by Li [9] , described in Section 2.4. As the electronic
bonding character of the reaction layer changes from insulating to metallic, electron
transfer across the C/M interface increases. This increased electronic interaction (a
long-range interaction of metallic-like characteristics) increases the resulting interfacial work of adhesion Wad . According to Eq. 2.4, the contact angle should decrease
as Wad increases. This seems to be supported by the experimental evidence presented
in Table 2.3.
It is important to note that in some active brazing applications the precipitation
of a reaction layer is not thermodynamically possible. Nevertheless, dissolution of
the ceramic substrate is promoted because of the strong anity between the active
element and the ceramic anion. Oxygen tends to segregate to metal free surfaces [19,
chap. XIV ], lowering LV in Eq. 2.4. Naidich [20] proposed that the dissolved oxygen and the active metal form clusters (because of the strong chemical interaction)
that have partially ionic character due to the charge transfer from the metal to the
oxygen atom. Due to this ionic character, these clusters can interact with ionicallybonded solid substrates and segregate to the solid/liquid interface. This interaction,
in turn, lowers SL in Eq. 2.4. If |SL | > |LV |, as has been observed in some systems (See Naidich [20]), then |Wad | > 0 and therefore || < 0. Wetting, thus, can
be improved in reactive brazing systems by just promoting further oxide dissolution
into the liquid braze.
The reason for this segregation is that the perturbations in the structure and energy of bulk
metals can be better relaxed in 2D geometries [5, chap. 6].
38
The eectiveness of active brazing in C/M joining applications is thus derived

from two main factors:
i) The active element can promote a greater ceramic dilution. Dissolved oxygen
reduces the Solid/Liquid interfacial energy and increases the work of adhesion of
the C/M system.
ii) If the interaction between the active element and the anion is strong, the precipitation of a reaction layer can be promoted. If this reaction layer has the
appropriate electronic properties wetting is further promoted and the work of
adhesion increases to levels that make C/M interfaces practical.
When the reaction product formed at the ceramic/liquid interface is ionically or
covalently-bonded, chances are that the liquid metals will not wet the substrate (or
de-wetting can occur). In this case, the use of active brazing alloys for the joining of
ceramic to metals will be ineective.
In order to accelerate the development of eective ceramic/metal joining techniques (in this particular case, brazing) it is desirable to be able to predict what
kind of interfacial interactions would occur in a particular ceramic/metal system. By
predicting the reaction products more likely to be observed under a particular thermochemical condition, it will be possible to assess the eectiveness of a particular
ceramic/metal joining approach. At the same time, with a sound knowledge of the
thermochemistry of ceramic/metal interaction, the interpretation and rationalization
of experimental results will be more straightforward.
2.7
Understanding Ceramic/Metal Interfacial Interactions: The need for

Thermodynamic Databases.
Because of their industrial importance, thermodynamic databases for material systems such as F e and N i-based alloys have reached a mature stage and the application
of such thermodynamic data to practical applications has become routine . Unfortunately, this cannot be said of many other material systems that, although important,
are either too recent in their developmental stage or have applications which are too
specic and hence they have not captured the interest of researchers dedicated to the
assessment and modelling of phase diagrams. In the particular case of ceramic/metal
joining applications, it is possible to identify key systems that require either further
study or a completely new analysis:
Active brazing alloys. As noted in Section 2.6, by using active chemical
additives it is possible to improve the wetting behavior of conventional brazing
alloys. Obviously, it is important to understand the thermochemistry of such
i.e. metallic in character.

For examples of such applications, please refer to Saunders and Miodownik [21, chapter 10]
39
alloys. Except for the work by Paulasto et al. [22], no attempt has been made
to develop a thermodynamic model for these systems .
Ceramic oxide systems. In order to model the interfacial interactions of
ceramic/metal systems it is necessary to have an accurate representation of
the thermodynamics of the ceramics of interest. Additionally, since many of
the possible reaction products are oxides themselves, it is essential to have as
complete a description as possible of such oxides. Although there is a lot of
experimental work on the study of interfacial reactions and reaction products,
almost no thermodynamic descriptions of these systems are available.
The purpose of this work is to provide a general methodology for the study of the
thermochemical interactions occurring at C/M interfaces. More specically, the focus of this thesis is the study of the interactions between zirconia-based ceramics and
metals, particularly during active brazing, although other cases are also studied. A
description of the Zr O system is fundamental. Because no satisfactory description
was available, a complete set of thermodynamic models for the principal phases of
this system was developed.
According to Section 2.4, the more stable the ceramic is, the less likely that there
will be a suciently strong work of adhesion Wad at the corresponding C/M interface.
From Fig. 2-3(a) it is obvious that ZrO2 is not likely to be wetted by most commercial brazing alloys. Active brazing alloys belonging to the Ag Cu T i family, for
example, should be used when attempting to join these stable ceramics to metals. A
thermodynamic description of the Ag Cu T i (and other braze systems such as the
Cu T i Zr, etc.) is therefore necessary for this work. Putting Cu T i alloys in
contact with ceramic oxides implies the possible presence of reaction layers belonging
to the complex Cu T i O family of oxides [23]. A thermodynamic description of
the Cu Zr O and T i Zr O systems is also of importance if a comprehensive
thermochemical analysis of all the possible interfacial reactions is to be achieved. As
will be shown later in this thesis, a correct thermodynamic description not only sheds
light on the nature of C/M chemical interactions but also can be used to gain a
deeper understanding of kinetics aspects of such interactions such as diusion paths,
interfacial morphology, etc.
In Appendix A, a general introduction to thermodynamic modelling is presented.
The two following chapters present a thermodynamic assessment of the metallic and
oxide systems of interest for this thesis. A detailed review of the available models
and experimental information will be presented. Discrepancies between the observed
and calculated thermodynamic properties of the systems will also be discussed.
In the paper by Paulasto et al. [22] the parameters of the thermodynamic models were not
reported.
40
Chapter 3
Development of Thermodynamic
Database for Active-Brazing Alloys
3.1
Development of Active Brazing Thermodynamic Database: Introduction
As noted in Section 2.7, very few attempts have been made at obtaining a selfconsistent and reliable thermodynamic database for active brazing alloys. Although
considerable research has been done regarding ceramic/metal active brazing techniques and processes, many of the key aspects of the resulting ceramic/metal interfaces, such as the reaction products formed, cannot be predicted, or even rationalized,
due to the lack of understanding of the thermochemical behavior of the active metal
brazing alloys used in those studies.
Even for the very important Ag Cu T i ternary system used in the majority
of ceramic/metal active brazing applications, thermodynamic models are incomplete.
The lack of a reliable thermodynamic description for this system makes it dicult to
design better brazing alloys. Because of this, an optimization of the brazing parameters for a particular ceramic/metal joining application cannot include active braze
composition as an optimizable parameter, unless one is willing to obtain (perhaps at
a great cost) a great number of experimental data points. It is therefore no surprise
that the actual active brazing alloys based in the Ag Cu T i system are limited
to a handful of compositions.
In addition to the widely used Ag Cu T i alloy system, novel brazing ller
metals based on the Cu T i Zr system have been developed for brazing of T ibased alloys [24]. Moreover, several workers have been exploring the use of such novel
ller metals in ceramic/metal joining applications [25]. The use of active brazing
alloys based on this system provides a high-temperature alternative to the lowertemperature active brazing alloys based on the Ag Cu T i system . Although the
use of brazing alloys based on the Cu T i Zr ternary system is not as extensive as
The melting range of the Cu T i Zr system (9000 C) is about 1000 C higher than the melting
range for widely used alloys belonging to the Ag Cu T i system.
41
that of Ag CuT i alloys, it is to be expected that its use will increase in importance
over time. It is therefore desirable to have a reliable thermodynamic model capable
of describing the thermochemical behavior of this system,
In the present chapter, thermodynamic models for both the Ag Cu T i and
Cu T i Zr active brazing systems will be developed using the often-called CALPHAD formalism presented in Appendix A. Given the fact that the complete procedure for the thermodynamic assessment of these two systems has been [26] or will be
made public in the near future, only the basics of the actual assessment procedures
and results will be presented.
The optimization of the thermodynamic parameters and the calculation of the resulting phase diagrams was done using the computational thermodynamics software
Thermo-Calc r . The parameters for all of the thermodynamic models described in
this chapter can be found in Appendix C.
3.2
Assessment of the Ag Cu T i System
As mentioned in Section 2.6, molten metals do not usually wet ceramic substrates , especially if the ceramics in question are stable, ionically-bonded oxides, such as Al2 O3 ,
ZrO2 , etc. When using liquid-based bonding procedures (such as brazing), wetting of
the substrates to be joined is of fundamental importance. In ceramic/metal brazing
applications, it is necessary to chemically modify the liquid brazing alloy so wetting
is promoted. Because of its chemical reactivity , T i is usually the additive of choice
when attempting to improve the wetting behavior of chemically-inactive brazing alloys.
The Ag Cu family of alloys is widely used in conventional brazing applications.
It was thus logical to attempt its use in ceramic/metal brazing applications. In order
to promote the wettability of these alloys, T i began being used as the reactive additive
and the Ag Cu T i system began to be extensively used in ceramic/metal joining
applications several decades ago. The low liquidus temperatures of the AgCu eutectic allows brazing at reasonably low temperatures, while the presence of Ag increases
the activity of T i very noticeably [27], promoting interfacial reactions with even the
most stable ceramics. Understanding the chemical behavior of the Ag Cu T i system is important when trying to understand the mechanisms of reactive wetting [28].
Although the Ag Cu T i system has been modelled before [22], no thermodynamic description has been made available. In this section, a model for the ternary
Ag Cu T i system will be proposed and phase diagram and thermochemical calculations will be compared with the experimental evidence available.
This is especially true when the molten metals in question do not have active element additions.
Titanium has a great anity for the anionic components of important industrial ceramic families
such as oxides, nitrides, carbides, etc.
Zr, Hf are also used, although on a much smaller scale.
42
3.2.1
Assessment of the Ag T i Subsystem
One of the greatest advantages of the CALPHAD methodology (see A) is that thermodynamic descriptions for low-order systems (binaries) can be easily expanded to
describe higher-order (ternary) systems by including higher-order interaction terms
and, when necessary, descriptions for ternary compounds [21]. In the case of the
Ag Cu T i system, by using the already developed thermodynamic descriptions
for Ag Cu [29], Cu T i [30] and Ag T i [31] it is possible to make the extensions
to the third-order Ag Cu T i ternary system with minimum eort, provided the
assessments of the binary subsystems are reliable. While the thermodynamic descriptions for the Ag Cu and Cu T i systems can be considered to be adequate , it was
necessary to carefully re-examine the description for the Ag T i binary since recent
results regarding the thermochemistry of Ag T i alloys [32, 33] contradict the early
thermodynamic assessment of this system [31].
3.2.1.1 Phases Present in the Ag T i System

According to the assessment by Murray and Bhansali [31], the Ag T i system is
comprised of the following phases:
The Liquid Phase: Murray and Bhansali [31] observed that the steep slope
of the liquidus near the T i and Ag-rich ends of the diagram, with an apparent attening in the central compositional range, hints at the existence of a
metastable liquid miscibility gap [31]. Although this miscibility gap does not
appear in the binary equilibrium phase diagram , its presence has been conrmed in experimental observations concerning the Ag Cu T i ternary (See
below). The existence of this metastable miscibility gap is a consequence of the
highly positive Ag T i chemical interaction and is therefore an indication of the
high T i chemical activity in Ag-rich melts. Thanks to this reactivity, Ag-based
brazing alloys need relatively small T i additions in order to make them reactive
enough so wetting of ceramic substrates is promoted.
The liquid phase can be modelled using the substitutional random sublattice
model, described in Appendix A, having at a least a regular solution parameter
so the positive interaction energy between Ag and T i can be represented with
the following notation:
(Ag, T i)1
The systems have been critically assessed and there have been no cases in which two sets of
experimental information were mutually excluding [29, 30].
The critical temperature for the liquid miscibility gap is lower than the solidus temperature
observed in the equilibrium phase diagram.
43
Solid Solution Phases: According to the assessment by Murray and Bhansali

[31], the solubility of T i in the f cc, Ag-rich solid solution is small, providing
further indication of a strongly positive Ag T i chemical interaction energy.
Although the solubility of Ag in bcc and hcp T i-rich solid solutions is higher,
it is reasonable to expect a similar positive interaction in these phases as well.
The solid solution phases, f cc, hcp and bcc can be modelled using the sublattice
formalism for an interstitial phase :
(Ag, T i)1 (V a)c
where c is the ratio between interstitial and normal sites in the solid solutions.
As noted above, c has a value of 0.5, 1 and 3 for the hcp, f cc and bcc structures,
respectively .
The AgT i Phase: The AgT i phase is iso-structural with CuT i. Experimental
evidence suggest that both phases present a certain degree of non-stoichiometry.
Because of their similarity, it was reasonable to expect that the model already
available to describe the CuT i phase [30] could be applied to the AgT i system.
In their assessment of the Cu T i system, Hari Kumar et al. [30] modelled
the CuT i phase considering that for the perfectly stoichiometric composition,
T i would occupy one sublattice, while Cu would occupy the other one. To
model the observed non-stoichiometric deviations it was supposed that point
defects (anti-structure atoms ) were also present in the phase . Using the
sublattice notation, this phase was represented by (Cu, T i)1 (T i, Cu)1 . Given
that the AgT i phase is iso-structural with CuT i, it is reasonable to use the
same sublattice formalism to describe its thermodynamic behavior:
(Ag, T i) (T i, Ag)
This allows the further modication of the model so interstitials such as O can be later introduced.
When considering substitutional solutions in these phases the actual site ratio between normal
sites to interstitial sites is not important. On the other hand, this ratio has a very physical significance when considering interstitial solutions. In the case of the f cc phase, there are one normal
site per interstitial site, while for the bcc phase, there is one normal site per three interstitial sites.
Although the hcp phase has a site ratio of 1 : 1, elastic interactions between interstitials eectively
reduce this ratio to 1 : 0.5 [34].
Anti-structure atoms are point defects corresponding to a complete inversion of the occupation
within the structure. For example, the anti-structure of CuT i4 is T iCu4 where the Cu atoms are
replaced by T i and viceversa.
This model is called the Wagner-Shottky model [35]. This formalism considers that an intermetallic phase can have point defects such as vacancies and anti-structure atoms and denes the
Gibbs free energy of such a phase using the Gibbs free energy of the perfect stoichiometric compound,
plus additions representing the Gibbs energy of formation of its point defects.
44
The AgT i2 Phase: There is no consistent experimental evidence regarding

the non-stoichiometry of the AgT i2 phase (it is iso-structural with M oSi2 ).
In their assessment, Murray and Bhansali [31] noted that this phase could be
considered as being perfectly stoichiometric. Therefore, a stoichiometric model
was used to describe this phase:
(Ag) (T i)2
3.2.1.2 Assessment Results
Using only phase diagram data it is possible to obtain approximate values for the
model parameters describing the thermodynamics of a system. The problem with this
approach, however, is that, in principle, there is a great number of correct solutions
to the thermodynamic optimization problem. It is desirable then to constrain the
problem in such a way that this optimization results in a realistic thermodynamic
description . Using thermochemical measurements (such as enthalpies, chemical activities, etc.) of any given phase, it is possible to unambiguously dene the parameters
describing its thermodynamic behavior. If there is experimental data on equilibria
between that phase and another one, the already optimized description can be used,
in conjunction with experimental data, to further restrict the possible values that the
parameters describing the second phase must have in order to obtain an optimal
description of the thermodynamics of the system.
Unfortunately, thermochemical information on the AgT i system is scarce and, as
will be seen later (see 3.2.1.3), in conict with well-accepted phase diagram features.
Therefore, in the preliminary re-assessment of the AgT i system, the thermochemical
measurements reported by Fitzner and Kleppa [32], Wei et al. [33] were ignored and
the optimization procedure was performed using only phase diagram data. In light
of the resulting optimized description, the thermochemical measurements reported
by Fitzner and Kleppa [32], Wei et al. [33] were critically examined and, as will be
seen below, were determined to be non-reliable.
As can be seen in Table 3.1 there is excellent agreement between the calculated
and experimental invariant points . Although agreement with experimental invariant
points does not necessarily guarantee the reliability of the model, it indicates that
the models used are at least a good rst-order approximation. In order to establish a higher degree of condence on the models developed, it would be necessary to
have additional independent data that could be used to further constrain the set of
parameters necessary to describe the system. In this particular case, experimental
phase boundary data involving equilibria between four dierent phases were used as
Phase boundaries, invariant reactions, etc.

A realistic description would represent both phase diagram topological features and thermochemical properties with a high degree of reliability. Additionally, such a description could be
condently used in both metastable and higher-order equilibria calculations.
Initial values for the parameters were taken from the model proposed by Kaufman [36]
45
Table 3.1: Comparison between calculated and experimental invariant points in the Ag T i
system.
Invariant Reaction
at.%Ag of Respective Phases
Temp.
0
C
1020 5
Exp.
91.5
49.5
1024
Calc.
48
33.33
940 5
Exp.
49
33.33
943
Calc.
7.6
4.7
33.33
855 5
Exp.
6.7
4.2
33.33
853
Calc.
95
50
95
959 1
Exp.
93.86
48
11.5
AgT i2
94
12
AgT i
bcc + AgT i
bcc
15.5
16.35
bcc + L
50
94.3
959
Calc.
hcp + AgT i2
AgT i + f cc
a further constraint to the thermodynamic optimization:

1100
1000
BCC
HCP+AgTi
Temperature, 0C
900
BCC+AgTi2
800
HCP+AgTi2
HCP
78Pli
63Rei
60McQ
700
600
500
0.05
0.10
0.15
0.20
x(Ag) Mole Fraction Ag
Figure 3-1: Experimental ( [3739] ) vs. calculated phase boundaries in Ag T i system.
Fig. 3-1 shows good agreement between calculated and experimental phase boundaries in the T i-rich region of the phase diagram . This agreement is encouraging, since
a correct prediction of the solubility limits for solid solution phases in equilibrium with
intermetallic phases (as in this case) provides additional corroboration for the chosen
parameters used to describe such compounds. It is necessary to emphasize, again,
The agreement is not perfect: there are conicting sets of experimental phase boundary data
points.
46
that information regarding the thermochemical behavior of this system would be fundamental to remove any remaining uncertainty regarding the ability of the models
used to correctly represent the thermodynamic behavior of the Ag T i system. As
will be seen in 3.2.1.3, such measurements have recently been made available.
According to the optimization results, the interaction parameters for all the solution phases in the Ag T i system are positive (See Appendix C). As will be seen in
Section 3.2.2, the liquid phase in the Ag Cu T i ternary system shows a miscibility
gap, which is caused, in part, by strong positive Ag T i and Ag CuT i interaction
energies. This result is also consistent with the high T i chemical activity in Ag melts
observed in many ceramic/metal brazing experiments [28] .
3.2.1.3
Problematic Thermochemical Measurements
In order to rene the above assessment, it was attempted to include recent thermochemical experimental information obtained by Fitzner and Kleppa [32]. Through
calorimetric measurements these authors determined the enthalpy of innitely diluted
T i in liquid Ag to be 52 kJ/mol. This constitutes a large negative deviation from
ideality and is contrary to the values obtained from the thermodynamic optimization
using phase diagram data (+25 kJ/mol). This negative contribution to the Gibbs
energy of the liquid phase would promote the existence of a deep eutectic at the Agrich side of the phase diagram, however, the accepted Ag T i phase diagram shows
a eutectic temperature of only 10 C below the melting point of Ag.
Further evidence against the measurements by Fitzner and Kleppa [32] comes from
analyzing the very similar Cu T i system. This system has far more intermetallic
compounds than the Ag T i system (6 vs. 2), so it is expected that the Gibbs
energy of the liquid phase would present a larger negative deviation from ideality .
From the assessment by Hari Kumar et al. [30], the enthalpy of liquid T i at innite dilution in liquid Cu is 9.95 kJ/mol. Using as an argument the fact that Ag
and Cu have very similar chemical properties, and considering the dierence in the
number of stable intermetallic compounds, it is possible to consider 9.95 kJ/mol as
a lower bound for the value of the enthalpy of liquid T i in liquid Ag at innite dilution.
Wei et al. [33] used the Electromotive Force Method, EMF, to measure the chemical activity of T i in Ag silver melts. They placed the liquid solutions in alumina
crucibles and observed the reaction products formed at the alumina/melt interface.
Because T i has a very high anity for oxygen, they determined the chemical activity
of T i by measuring the equilibrium oxygen partial pressure in the melt, considering
the following reaction:
Evidence for this high chemical activity comes from the nature of reaction products observed
in ceramic/metal joints using brazing alloys based on the Ag T i system
In general, the presence of a large number of intermetallic phases indicates a negative deviation
from ideality in the liquid phase. See Bakker [40] for a more comprehensive discussion on this
observed trend.
47
1
xT i + O2 (g) = T ix O(s)
(3.1)
2
where x denes the stoichiometry of the titanium oxide present, and the T i represents
titanium dissolved in the Ag melt.
From Eq. 3.1, it can be seen that the equilibrium oxygen partial pressure is determined by the chemical activity of T i in the melt and the titanium oxide in equilibrium
with the Ag T i liquid:
1
Keq = x
(3.2)
aT i (pO2 )1/2
One of the key parameters used in the calculation of the chemical activity of T i was
the T iO oxide phase in equilibrium with the melt. Wei et al. [33] used XRD to identify the phase and were surprised to nd T i2 O (essentially, hcp T i saturated with O
interstitials) as the equilibrium phase. This phase, however, has been observed to be
present when the activity levels of T i are very high [28]. Despite this evidence, the
authors interpreted the observed phase as the result of a solid-state transformation
from T iO to T i2 O.
Extrapolating the mathematical tting of the measured data down to innite dilution, Wei et al. [33] obtained an enthalpy of Ti at innite dilution in liquid Ag of
27, 850 J/mol. Again, a negative deviation from ideality is observed, although this
deviation is considerably less than the one obtained by Fitzner and Kleppa [32]. This
value, however, cannot be reconciled with the features of the accepted experimental
Ag T i phase diagram.
It may be possible to question the results by Wei et al. [33] in light of the
misidentication of the oxide (T iOx vs. T i2 O) formed at the crucible/melt interface.
By using the Gibbs energy of formation for T i2 O reported by Pak et al. [28], and,
taking the measured values for the oxygen partial pressures as the correct ones, an
increase of an order of magnitude in the values for the activities of T i in Ag melts
can be obtained. This new interpretation of the work by Wei et al. [33], however, is
not enough to reconcile these measurements with the accepted features of both the
Ag T i and Ag Cu T i phase diagrams, and the apparent high chemical activity
of T i when dissolved in Ag melts.
In their study on the thermochemical behavior of the Ag Cu T i liquid alloys,
Pak et al. [28] used an EMF method similar to that used by Wei et al. [33] to obtain
the chemical activity of T i in Ag Cu melts. Although Pak et al. [28] focused their
investigations of the thermochemistry of T i in Ag Cu melts, they report a single
measurement of the chemical activity of T i in a Ag liquid solution with no Cu. The
measurement cannot be extrapolated to innite dilution(there is only one data point),
but it clearly shows a very positive deviation from ideality. This result contradicts the
results obtained by Wei et al. [33] but is consistent with the accepted experimental
phase diagram.
48
1800
1600
Liquid
BCC
1200
800
HCP
600
0.2
AgTi
1000
AgTi2
Temperature, 0C
1400
0.4
0.6
Mole Fraction Ag
FCC
0.8
1.0
Figure 3-2: Calculated AgT i phase diagram. Triangles indicate experimental invariant
points.
Although there are enough reasons to question the results obtained by Fitzner
and Kleppa [32] and Wei et al. [33], the fact that two independent and un-related
thermochemical measurements oered results contrary to what would be expected
according to accepted phase diagram features makes it impossible to reject them
completely. A sensitivity analysis on the eect of varying the values of the interaction parameters for the liquid and fcc phases on the compositional coordinate of the
reaction L
AgT i + f cc (xing the reaction temperature at the accepted 9600 C),
shows that as the binary chemical interaction parameter on both phases became less
positive (going towards the values proposed by [32, 33]), the eutectic composition
was displaced towards the T i rich side of the diagram. With a negative interaction
coecient, a eutectic point was present at the middle of the phase diagram, and the
width of the bcc T i single-phase eld was decreased to values much lower than the
accepted solubility limits. Using this sensitivity analysis, along with the independent
data by Pak et al. [28], it was decided that the phase diagram measurements were to
be accepted, while the data by Fitzner and Kleppa [32], Wei et al. [33] were to be discarded, at least until more reliable thermochemical measurements are made available.
This section has shown how the CALPHAD approach, when followed properly,
constitutes a self-consistent method that can be used to discriminate between conicting pieces of experimental evidence. Thus, the CALPHAD approach not only
provides a systematic way to model the thermodynamic properties of material phases
but also allows one to determine the reliability of the experimental data used to construct such models.
49
3.2.2
Assessment of the Ag Cu T i Ternary
Once the reliability of the thermodynamic models of the Ag T i, Ag Cu and

CuT i sub-systems has been established, a thermodynamic description of the ternary
Ag Cu T i system can be obtained by using available experimental information.
3.2.2.1
Phases Present in the Ag Cu T i System
Eremenko et al. [41, 42, 43, 44] assessed the Ag Cu T i system at several temperatures and over the entire compositional range. From their work, the following
condensed phases can be identied :
Liquid Phase: The liquid phase can be modelled using a regular random solution single sublattice model. This phase was observed [41] to undergo phase
separation in the central region of the phase diagram. Since this liquid miscibility gap was not observed in the binary sub-systems, they assumed that the
miscibility gap would have a closed topology. The model can be represented by:
(Ag, Cu, T i)
M T i2 Phase: Eremenko et al. [42] concluded that the CuT i2 and AgT i2 phases
were iso-structural (M oSi2 is their prototype). At the xCu = 0.33 composition
and low temperatures, they found that the lattice parameters changed continuously from CuT i2 to AgT i2 , implying complete solid solubility among the two
phases. It was natural, then, to model the two phases as if they were a single
phase. Using the sublattice formalism, this phase can be modelled as:
(Cu, Ag)1 (T i)2
This model allows random mixing of Cu and Ag atoms in the rst sublattice, as
evident from the experimental results by Eremenko et al. [43, 44] and Paulasto
et al. [22].
M T i Phase: The CuT i and AgT i phases are iso-structural [42]. Nevertheless,
they do not form solid solutions and are not even in equilibrium with each other
at any temperature, although both phases exhibit noticeable penetration in the
ternary compositional triangle [22, 42]. Therefore, it was decided to represent
them as a single phase, using the model proposed by Hari Kumar et al. [30],
allowing for mixing of all three atoms in the two sublattices:
(Cu, Ag, T i)1 (Cu, Ag, T i)1
Only the phases actually assessed during this work are included in this description
50
3.2.2.2 Experimental Data used in the Assessment of the Ag Cu T i Ternary

In a series of papers spanning two years, Eremenko et al. [41, 42, 43, 44] generated
the most important collection of phase diagram data on the Ag Cu T i system at
dierent temperatures to date. Through their work, they discovered a miscibility gap
in the liquid phase [41] (not observed in any of the binary sub-systems), the phase
relationships of the ternary section at 7000 C, along with the liquidus surface and a
series of invariant ternary reactions involving the liquid phase [43].
Paulasto et al. [22] used the diusion couple method to study the Ag Cu T i
system at 9000 C, 9500 C and 10000 C and at dierent compositions. In these series
of experiments, Paulasto et al. [22], observed the formation of two liquid phases with
dierent densities, conrming the results by Eremenko et al. [41]. At 9500 C, extensive
penetration of the CuT i phase into the ternary composition triangle was observed.
Pak et al. [28] used the EMF Method to measure the chemical activity of T i
in CuAg eutectic melts. They placed the liquid solutions in alumina and zirconia
crucibles and observed the reaction products formed at the crucible/melt interface.
Because T i has a very high anity for oxygen, they determined the chemical activity
of T i by measuring the equilibrium oxygen partial pressure in the melt using the
following equilibrium reaction:
1
xT i + O2 (g) = T ix O(s)
2
(3.3)
where x denes the stoichiometry of the titanium oxide present, and the underlined
T i represents titanium dissolved in the CuAg melt.
3.2.2.3
Optimization of the Ag Cu T i System
In order to assign values to the parameters of the models outlined in 3.2.2.1, the phase
diagram, invariant equilibria, diusion and electrochemical data briey described in
3.2.2.2 were used. Existing assessments of the Ag Cu [29] and Cu T i [30] systems
were also incorporated into the optimization.
Fig. 3-4 presents the calculated Ag Cu T i phase diagram, along with experimental data points obtained from the literature (See [22, 42]). In general, the
agreement is quite good:
i) Fig. 3-4(a) shows the calculated and experimental Ag Cu T i phase diagram
at 7000 C. As can be seen, in most of the compositional triangle the agreement
is excellent and the single, two- and three-phase elds have adequate compositional ranges and all the important features of the experimental diagram are
Electromotive force measurements consist of measuring the chemical potential of an anionic

species through the measurement of the electric potential across a membrane capable of maintaining
a gradient of chemical potential of such anion across its thickness.
51
1100
2000
1000
1600
800
Temperature, 0C
1400
700
600
1200
fcc#1+fcc#2
500
0.2
0.4
0.6
Mole Fraction, X(Ag)
0.8
1.0
(a) Calculated Ag Cu system, after

model by Hayes et al. [29].
600
0.2
0.4
0.6
Mole Fraction Ti
CuTi2
CuTi
Cu2Ti
Cu4Ti3
800
300
Cu3Ti2
1000
400
200
BCC
Liq
FCC
Cu4Ti
Temperature, 0C
1800
Liquid
900
HCP
0.8
1.0
(b) Calculated Cu T i system, after

model by Hari Kumar et al. [30]
Figure 3-3: Calculated subsystems of the Ag Cu T i ternary system.
reproduced.
ii) Fig. 3-4(b) shows the experimental and calculated ternary Ag Cu T i phase
diagram at 10000 C. As can be seen, the liquid bcc two-phase equilibrium is
well-reproduced. This two-phase equilibrium was also reported by Eremenko
et al. [41], although they did not report the exact compositions of the phase
boundaries.The direction of the tie-lines indicating regions across the miscibility
gap at thermochemical equilibrium also correspond to reported experimental
results [41].
As noted above, Eremenko et al. [43] used metallographic and thermal analysis
techniques to determine the existence of three invariant transformations taking place
on alloys within the Ag Cu T i system during heating or cooling. These three
invariant transformations occur when the three simultaneous crystallization curves
for the phases T i and CuT i2 ; CuT i2 and CuT i; CuT i and Cu4 T i3 starting from
the L
T i + CuT i2 , L
CuT i2 + CuT i and L + CuT i
Cu4 T i3 binary invariant
points meet the boundary of the miscibility gap region of the ternary system.
As can be seen in Table 3.2, the agreement between experimental and calculated
invariant points is good. There are some discrepancies, however, in the temperatures
for the invariant reactions. The largest discrepancy is for the invariant reaction L#1+
CuT i
L#2+Cu4 T i3 . Since the thermodynamic properties of the CuT i and Cu4 T i3
phases are already dened and veried trough the assessment of the binary CuT i by
Hari Kumar et al. [30], the only possible explanation for this discrepancy lies in the
ternary interaction parameter used to describe the Gibbs energy of the liquid phase.
The fact that the calculated temperature for all the invariant reactions is lower than
the experimental temperatures, indicates that the liquid phase might be too stable,
52
0.7
0.4
0.3
Liq#1+Liq#2
0.7
0.6
CuTi
0.4
0.3
0.2
Cu4Ti
0.2
0.8
MTi
Cu4Ti3
Cu3Ti2
MTi
rac
le F
rac
0.6
tio
MTi2
Mo
tio
nX
(Ti
0.8
BCC
0.9
1 Phase
2 Phase
3 Phase
(Ti
0.9
Mo
le F
1.0
HCP
nX
1.0
0.1
0.1
0 FCC 0.2
0.4
0.6
0.8 FCC 1.0
Mole Fraction X(Cu)
(a) Calculated and experimental phase

diagram of the Ag Cu T i system at
7000 C. The experimental data was obtained from [42].
Liquid
0.2
0.4
0.6
0.8
1.0
Mole Fraction X(Cu)
FCC
(b) Calculated and experimental phase

diagram for the Ag Cu T i system at
10000 C. Experimental data after [22].
Figure 3-4: Calculated phase diagram of the Ag Cu T i at two dierent temperatures.

Experimental data points are incorporated for comparison.
and a more positive deviation from ideality might correct the discrepancy. From a
sensitivity analysis on the ternary interaction parameter of the liquid phase it was
found that, when the interaction parameter had values above 40, 000 J/mol it was
impossible to nd equilibria involving the invariant reactions presented in Table 3.2.
It is possible, however, that this problem could be resolved if the ternary interaction
parameter for the liquid phase were implemented in its asymmetrical form (please
refer to Appendix A):
Table 3.2: Experimental [43] and calculated invariant points in the ternary Ag Cu T i
system
Invariant Reaction
Temp.0 C
xCu
xTi
exp.
calc.
exp.
calc.
exp.
calc.
L#2 + CuT i2
980
996
0.25
0.22
0.6
0.55
L#2 + CuT i2 + CuT i
954
991
0.4
0.27
0.55
0.52
900
837
0.62
0.57
0.37
0.38
L#1 + T i
L#1
L#1 + CuT i
L#2 + Cu4 T i3
53
L =
i,j,k
x L
i i,j,k
(3.4)
where L is the ternary interaction parameter between elements i, j, k in substitui,j,k

tional phase .
The problem with this approach, however, is that more degrees of freedom are
introduced to the thermodynamic optimization, and the numerical values used to t
the experimental results could lead to anomalous extrapolations in regions away from
the compositions explored by the experimental work. Additionally, there is also the
possibility of experimental inconsistencies due to the fact that liquid alloys involving
the Ag Cu T i system are likely to undergo extensive Ag evaporation at high
temperatures [31]. Thus, since the reliability of the experiments could not be fully established, it was judged appropriate to consider the calculated results as good enough
for the purposes of this work.
0.2
Exp
Calc
Activity Ti. Ref. State: BCC T=1000
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0.005
0.01
0.015
0.02
0.025
0.03
Mole Fraction X(Ti)
Figure 3-5: Calculated and experimental chemical activity of T i in the eutectic melt AgCu
at 10000 C. The experimental data points are taken from [28]
3.2.3
Summary
In ceramic/metal joining applications that use active brazing, the chemical activity
of T i in the liquid brazing alloy is perhaps the most important factor when trying
to predict the reaction products formed at the ceramic/metal interface. Thus, a
reasonable calculation of the chemical activity of T i in liquid Ag Cu T i melts
is desired. Fig. 3-5 shows the comparison between experimental and calculated T i
chemical activities in eutectic Ag Cu melts. As can be seen, the agreement is quite
good. It is important to note, however, that the calculated values are higher than the
54
experimental ones. This implies a higher calculated positive deviation from ideality
than experimentally observed.
The correct calculation of the ternary Ag Cu T i system contributes to the
verication of the reliability of the parameters used for the binary Ag T i system.
This is a signicant result, since many inconsistencies between calculation and experiments have become evident in the analysis of this binary system (See 3.2.1). Since
the Ag Cu T i system is extensively used in ceramic/metal joining applications,
the results from this work will contribute to better understanding and optimization
of technologically relevant ceramic/metal interface engineering.
3.3
Assessment of the Cu T i Zr System
The assessment of the AgCuT i presented in Section 3.2 is key to understanding of

the thermochemical behavior of an important family of active brazing alloys [22]. Despite the successful use of this brazing alloy, a need for higher-melting active brazing
alloys has recently emerged, especially in ceramic/metal joining applications involving
zirconia-based ceramics. Such brazing alloys are based on novel amorphous metallic
systems based on the CuT iZr ternary. It is therefore desirable to obtain a reliable
description of this system that could not only be used to design better active brazing
alloys but also would aid the development of novel amorphous alloys, as stated in [26].
3.3.1
Description of the Cu T i Zr System
Woychik and Massalski [45] studied the isothermal section of the Cu T i Zr system at 7000 C. Microstructures of the resulting samples were analyzed using optical
microscopy, while their chemical composition was determined using electron microprobe analysis. The phase relationships at 7000 C were determined by tting of the
corresponding three-phase triangles.
One of the dominant features of the 7000 C section was the presence of the ternary
phase Cu2 T iZr. This phase greatly inuences the shape of the liquidus at the central
region of the phase diagram and is expected to be an important competing crystalline
phase in amorphous solidication processing involving this system. It was also observed that this phase enters into pseudo-binary equilibrium with most of the stable
phases of the system at the experimental temperatures.
To determine any possible eutectic reactions involving this phase, Woychik and
Massalski [45] used the Transient Liquid Phase (TLP ) bonding technique [46]. In
this procedure, two phases, A and B, are brought into contact and assembled into a
sandwich structure- A/B/A, for example- at a temperature higher than the invariant
point involving these phases. If there were a eutectic-type reaction involving these
two phases, inter-diusion processes would eventually lead to the formation of a liquid layer at the A/B interfaces. A liquid thus formed, gradually disappears as the
55
inter-diusion process continues. However, if the liquid is quenched before it disappears, the eutectic regions can be identied in the quenched microstructure and
the composition of the liquid at the eutectic point can be determined. Woychik and
Massalski [45] used diusion couples involving Cu2 T iZr and either CuT i, CuZr or
CuZr2 . From these experiments, the three eutectic reactions involving the Cu2 T iZr
phase were determined.
Chebotnikov and Molokanov [47] studied the CuT i2 -CuZr2 vertical region of the
CuT iZr system by preparing several alloys along the xCu = 0.33 iso-compositional
line. XRD analyses were performed both at high (297-697 0 C) temperatures and at
room temperature. Furthermore, DTA analysis was done on each of the samples for
the entire temperature range until complete melting was detected.
XRD analyses showed that the alloys had linearly increasing a and c unit-cell parameters -while the c/a ratio decreased- as the amount of Zr increased over the entire
compositional range. This indicated that the alloys formed a single phase in which
Zr and T i atoms could be exchanged within the same lattice site of the unit cell. In
these XRD analyses, small amounts of a ternary phase (of un-reported composition)
in the central region of this iso-compositional section were detected. This ternary
phase may correspond to the Cu2 T iZr phase observed by Woychik and Massalski
[45].
3.3.2
Phases Described in the Cu T i Zr System
According to the experimental studies presented in 3.3.1 and the previous assessments
for the subsystems CuT i, CuZr and T iZr [30, 48, 49], the main phases present
in this system are:
Solution Phases: The liquid phase was modelled using a single-lattice random
solution model. Since all the elements comprising this system form substitutional fcc, hcp and bcc solutions, they could also be modelled using the same
single-lattice random solution model used for the liquid phase:
(Cu, T i, Zr)
However, if one is to allow further modications of the solid solutions (allowing
the incorporation of interstitials into the solid phases) it is necessary to use a
two-sublattice model:
(Cu, T i, Zr)1 (V a)c
where c is 0.5, 1 or 3 for the hcp, bcc and f cc phases, respectively.
The models describing the thermodynamic behavior of the solid solution phases
f cc, and hcp were not re-assessed. The liquid and bcc phases, however, were
further modied by adding an extra term to the expression for ex G (See 198).
m
CuM2 Phase: In their experimental work on the CuT iZr system Chebot56
nikov and Molokanov [47] determined that the phases CuT i2 and CuZr2 form
a continuous series of solid solutions across the entire xT i + xzr = 2/3 isocompositional line. Therefore, it was decided to model this phase as a single phase using a two-sublattice model. Cu would occupy the rst sublattice
while T i and Zr atoms formed a random solution within the second sublattice.
Accordingly, the sublattice notation for this phase is:
(Cu) (T i, Zr)2
Cu2 T iZr Phase: In their study of the isothermal section of the Cu T i Zr

system, Woychik and Massalski [45] identied the ternary Cu2 T iZr Laves
phase in the central region of the phase diagram. Since no information on
its solubility limits is available, it was decided to represent this phase as a
stoichiometric compound:
(Cu)2 (T i)1 (Zr)1
3.3.3
Optimization of the Cu T i Zr System: Results

FCC
1.0
0.9
Cu4Ti
0.7
Cu3Ti2
Cu4Ti3
CuTi
Cu5Zr
Cu51Zr14
Cu8Zr3
Cu10Zr7
u
at.
C
%
Two-Phase Equilibrium
Three-Phase Equilibrium
Cu2TiZr
CuTi2
CuTi2
0.3
0.2
0.1
HCP
0.2
BCC
HCP
0.4
0.6
0.8
1.0
% at. Zr
Figure 3-6: Experimental vs. calculated Cu T i Zr isothermal section at 7030 C
Fig. 3-6 shows the calculated and experimental [45] ternary section for the Cu
T i Zr system at 7030 C. The agreement in most of the phase diagram is good,
although there are some important discrepancies. First, Woychik and Massalski
[45] have established the existence of a two-phase equilibrium phase eld between
the phases Cu3 Zr2 and Cu3 T i2 . It was found that it was not possible to reproduce
this feature of the phase diagram without modifying the descriptions for the Gibbs
57
energies of formation for the binary phases in the Cu T i and Cu Zr systems, as

obtained by Hari Kumar et al. [30], Zeng et al. [48]. The other important discrepancy
is the reported existence of a three-phase eld involving the phases hcp, CuZr2 and
Cu2 T iZr. This contradicts the experiments done by Chebotnikov and Molokanov
[47].
One of the most important kinds of experimental data in terms of its usefulness
for thermodynamic modelling is the determination of invariant equilibrium points.
The consistency between experimental and calculated invariant equilibria constitutes
a good indicator of the validity of the thermodynamic descriptions used. Table 3.3
presents a comparison between the calculated and experimental invariant points, including reaction temperatures. One of the most satisfactory results was the successful
calculation of the temperatures for the invariant reactions. From TLP experiments
it is not possible to determine the temperatures at which the invariant reaction occurred. Nevertheless, all the TLP experiments occurred at temperatures between
8550 C and 8670 C. In order to observe eutectic (or peritectic) transformations upon
quenching of the TLP diusion couples, these invariant reactions must have occurred
at temperatures below that temperature range. Table 3.3 shows that this is, indeed,
what has been calculated.
Table 3.3: Comparison between calculated and experimental invariant points in the Cu
T i Zr system.
Invariant
Reaction
at.%Cu
at.%Ag
exp
calc
exp
calc
Temp.
0
C
L CuZr + CuZr2 + Cu2 T iZr
48.6
44.5
37.5
41.2
833
L bcc + CuZr2 + Cu2 T iZr
39.4
38.5
43.2
29.5
827
L + CuT i2 CuT i + Cu2 T iZr
47.6
44.9
17.8
14.5
856
Fig. 3-7 shows the calculated and experimental [47] pseudo-binary section CuT i2
CuZr2 . As can be seen, very good agreement has been achieved between the points
describing the solidus and liquidus lines and the calculated pseudo-phase boundaries.
According to the experiments, a minimum in the liquidus temperature is reached in
the central region of the phase diagram (33 at. %) calculated vs. 36.7 at. % experimental). At this point, the liquid phase is stable down to 8390 C, which compares well
with the calculated value of 8270 C. It can also be seen that the calculation shows
that a continuous series of compositions based on the (Cu)1 (T i, Zr)2 formula exist at
temperatures below 8270 C. The stability of the CuM2 phase over the entire compositional range from CuT i2 to CuZr2 has been veried experimentally. Chebotnikov
and Molokanov [47] observed the presence of a ternary phase (they did not identify
it) near the minimum of the liquidus in the pseudo-binary section. Fig. 3-7 shows two
58
33% Cu, Fixed

1027
BCC+Liq
Temperature, 0C
977
BCC+Liq+CuTi2
927
Liq+CuTi2
877
Liq
827
CuTi2+CuTi
CuTi2+CuZr
CuTi2+CuTi+HCP
777
CuTi2+Cu2TiZr
727
0.1
0.2
0.3 0.4 0.5
Mole Fraction Zr
0.6 0.6667
Figure 3-7: Calculated and experimental pseudo-binary section CuT i2 CuZr2 .
invariant reactions, close to 8270 C that involve the presence of the Cu2 T iZr phase.
3.3.4
Summary
The Cu T i Zr system was thermodynamically assessed in order to obtain an optimized and consistent thermodynamic description of all the experimentally-observed
stable phases. A new description for the phases CuT i2 and CuZr2 was proposed.
Isothermal sections and the surface for primary solidication were calculated using
the assessed models. It was found that the liquid is more stable at the central region
of the phase diagram.
59
60
Chapter 4
Development of Thermodynamic
Database for Ceramic Oxide
Substrates and Reactive Brazing
Interfacial Products
4.1
CALPHAD Modelling of Ceramic Oxide Systems: Introduction
As noted in Section 2.7, reliable thermodynamic models of ceramic substrates are

needed in order to obtain a better understanding of the underlying thermochemical
processes occurring at ceramic/metal, C/M interfaces. Additionally, understanding
the energetics of the likely reaction products resulting from interfacial C/M interactions would allow a better prediction of the equilibrium (both stable and metastable)
states of the C/M interfaces created through joining operations. Although there have
been attempts to describe the thermodynamics of some important ceramic/metal interaction systems through computational thermodynamics techniques [50], much work
remains to be done, especially when attempting to study ceramic/metal systems that
have not been broadly investigated even from the experimental point of view.
Over the past years, ceramic/metal joining operations involving the technologicallyimportant zirconia-based ceramics have become increasingly important due to emerging technologies that rely on the creation of stable zirconia/metal interfaces [51]. A
reliable thermodynamic model of such an important ceramic system is therefore essential for a better understanding of interfacial interactions between zirconia-based
substrates and metallic systems. Even though this class of ceramics have been studied
for several decades, there have been very few attempts at modelling their thermodynamic properties over the entire range of compositions and temperatures [52]. As a
part of this chapter, the essential aspects of the thermodynamic model for the Zr O
are presented, based on the much more extensive description presented in Arroyave
et al. [52].
Besides requiring accurate descriptions of the thermodynamic behavior of ceramic
61
substrate materials, a reliable description of the thermochemical behavior of likely

ceramic/metal reaction products (in this case oxides) is necessary in order to understand (and even have predictive capabilities) of the interfacial processes occurring
during ceramic/metal interactions. In the case of active-metal joining of ceramic oxide substrates to metals, a complete description of the Cu T i O, Cu Zr O
and T i Zr O systems is essential for the understanding of the reaction sequences
observed in the corresponding ceramic/metal interfaces.
Fortunately, as will be seen below, the relevant regions of the ternary systems
in question have been experimentally accessed, making the modelling of the thermodynamic properties of these systems possible. For example, the Cu T i O
system has been studied in the region where the M6 X compounds are stable. This
region is of great importance in active metal joining operations because the usual
reaction products observed belong to this M6 X family of compounds. In the case of
the Cu Zr O system, there is only one published paper on the thermochemical
properties of Cu Zr O melts at low oxygen concentrations. Although this single
experimental work could, in principle, be used for the thermodynamic optimization
of the liquid phase in this region of the diagram, it will be shown that these results
are not reliable. The T i Zr O system is the most extensively investigated ternary
of this set and it is therefore reasonable to expect that its assessment will be the most
complete. It will be shown that this is indeed the case, allowing some unimportant
discrepancies in regions of the system that have no direct impact on this thesis.
The optimization of the thermodynamic parameters and the calculation of the
resulting phase diagrams was done using the computational thermodynamics software Thermo-Calc r . For an explanation of the models used to describe the systems
described in this chapter, please refer to Appendix A. The parameters for all of the
thermodynamic models described in this chapter can be found in Appendix C.
4.2
Assessment of the Zr O System
Understanding of the chemical behavior of the Zr O system is fundamental for a

wide range of relevant industrial applications, such as glass-forming metallic alloys,
SOFC, ceramic/metal composites, thermal barrier coatings in turbine blades and so
forth (see Section 5.2).
Notwithstanding its practical importance, the only thermodynamic model available for the elemental Zr O system up to now was developed by Kaufman and
Clougherty [53] almost four decades ago. Since the publication of that model, considerably more experimental data has been obtained and therefore it is now possible
to obtain a more accurate description of the thermodynamic behavior of such an important system.
This is quite surprising, given the importance that this system has on several technologies
developed in the past few years.
62
In this section, the thermodynamic model for the Zr O system will be presented.
The models obtained from this work will be used in the thermochemical modelling of
ceramic/metal interfacial reactions.
4.2.1
Phases present in the Zr O System
The most complete critical assessment of the phase diagram information on this system corresponds to the work by Abriata and Versaci [54]. The Zr O system is of
medium complexity, having several condensed-mixture phases, stoichiometric compounds and ordered-interstitial solutions:
Gas Phase: In order to calculate the oxygen-rich side of the Zr O phase
diagram, the gas phase had to be considered . The ideal gas model was used,
considering the following species:
(O, O2 , O3 , Zr, Zr2 , ZrO, ZrO2 )
Liquid Phase: The liquid phase goes from pure liquid zirconium to stoichiometric ionic liquid ZrO2 . Therefore, any model used would have to be able to
represent the Gibbs free energy of the liquid phase from pure liquid zirconium
to pure stoichiometric ZrO2 in a continuous manner. The ionic two-sublattice
model for liquids [55] has been found to be extremely useful for this purpose:
Zr+4
42yO2
O2 , V a4
Solid Solutions: It is well-known that the stable solid phases of zirconium (bcc
and hcp) dissolve oxygen interstitially into their octahedral interstitial sites [56].
The solid solutions of Zr O can be represented with the two-sublattice formalism, with one sublattice occupied by zirconium atoms and the other one
occupied by both oxygen and vacancies:
(Zr)1 (O, V a)c
where c corresponds to the ratio of interstitial sites to normal sites in each
structure.
For the bcc phase, the stoichiometry ratio c is equal to 3. For the hcp phase,
the stoichiometry ratio used was 1/2 .
Ordered Interstitial Solutions: In binary interstitial alloys, the solubility of
interstitial elements into metals is usually so low that interstitial ordering is not
Solid-gas equilibrium calculations were necessary to optimize the properties of the ZrO2 solid
phases.
Although in principle it is possible for the interstitial oxygen atoms to occupy twice as many
vacant cites, strain energy eects make the existence of rst nearest-neighbor oxygen interstitials in
hcp structures highly unlikely [57]
63
observed. However, the hcp phases of IV A transition elements dissolve large

amounts of oxygen in their octahedral interstitial sites. In all these systems, a
tendency for ordering has been observed in the low temperature, high oxygen
content region of the solid-solution elds [58]. For this work, the phases Zr6 O,
Zr3 O and Zr2 O were modelled as stoichiometric compounds,
(Zr)c (O)
The ZrO2x Phases: The ZrO2x compound can exist, at 1 atm., in three
dierent structural modications. The low-temperature phase, ZrO2x , has
a monoclinic structure and has been reported to exhibit a certain degree of
non-stoichiometry. The medium-temperature tetragonal phase, ZrO2x (prototype: HgI2 ), has a higher degree of non-stoichiometry. In these phases, it has
been observed that under reducing conditions, oxygen vacancies are the predominant defects[59, 60]. Although there is information regarding phase boundaries
and thermochemical properties of the high-temperature cubic phase, ZrO2x
(prototype: CaF2 ), there is virtually no information on the defect structure of
this phase. This phase presents a large degree of non-stoichiometry and is stable at temperatures far lower than the transformation temperature for
the stoichiometric phases. There are however, some other non-stoichiometric
oxide systems that present large degrees of non-stoichiometry and for which
a defect structure has been established, CeO2x being one example [61]. For
these other oxide phases, vacancies have been used to successfully model their
thermochemical properties under reducing conditions. Therefore, it is possible
to consider that for the ZrO2x phase, vacancies can be considered to be the
primary defect.
For these three phases, the formation of defects can be understood using the
following reaction:
1
2
Olattice
O2,g + V a + 2e
2
According to this equation, the reduction of the zirconia phase involves the
formation of an oxygen vacancy plus two free electrons. In fact, experiments
have shown that under reducing conditions, all the allotropes of the zirconia
phase change their color from white to gray or black, indicating the presence of
conduction electrons [62]. Despite this fact, it has been shown that a satisfactory
description of the non-stoichiometry and its relationship to oxygen potentials
is possible without considering conduction electrons [57]. In their work on
the T i O system, Waldner and Eriksson [57] used a two sublattice model
to describe the uorite-type T iO2x phase. In this work, we use the same
description, giving the oxygen vacancy a double negative charge to maintain
electro-neutrality of the phase:
Zr+4
64
O2 , V a2
Table 4.1: Comparison between calculated and experimental invariant points in the Zr O
system.
Reaction
L + (hcp)
L
10.0 0.5
10.5 0.5
1970 10
Exp.
22.17
10.5
1977
Calc.
40 2
35 1
62 1
2062 10
Exp.
41.22
31.1
62.54
2074
Calc.
63.6 0.4
31.2 0.5
66.5 0.1
1525
Exp.
63.49
30.41
66.5
1530
Calc.
25.0 2
2130 10
Exp.
27.5
ZrO2x
hcp + ZrO2x
19.5 2
10.03
bcc
hcp + ZrO2x
2127.8
Calc.
hcp
4.2.2
Temp. 0 C
at.%O of Respective Phases
Model Optimization and Comparison with Experimental Results
The parameters needed for the description of the Gibbs Free Energy expressions
describing all the phases present (except for Gas phase) were optimized using the
PARROT subroutine of the Thermo-Calc r software mentioned in Section A.1. A
detailed account of the optimization is presented in Arroyave et al. [52].
4.2.2.1 Comparison with Phase Diagram Data
2550
Liquid
2350
Temperature, oC
ZrO2-x
BCC
2150
Ack 77
Ack 78
Geb 61
Rau 80
1950
1750
1550
ZrO2-x
HCP
1350
1150
ZrO2-x
0
0.2
0.4
0.6
0.8
Oxygen Mole Fraction X(O)
1.0
Figure 4-1: Experimental data vs. calculated Zr O phase diagram. Experimental data
obtained from Ackerman et al. [63, 64], Rauh and Garg [65], Gebhardt et al. [66]
As can be seen in Fig. 4-1 and Table 4.1, the experimental and calculated phase
65
diagram data agree well, except for some minor discrepancies:

The bcc/(hcp + bcc) phase boundary calculated in this work presents an oxygen
solubility slightly lower than that presented in the literature [66].
One of the major discrepancies between the experimental information and the
calculated phase diagram is the maximum solubility of oxygen within the hcp
phase (35 at.% vs.31 at.%).
Another major discrepancy is the composition at which the hcp melts congruently. The calculated value is 2at.% higher than that reported in the literature.
In similar transition metal-oxide systems, namely T i O, this composition is
close to the maximum solubility composition [57], and thus there are reasons to
believe that the congruent melting composition should be on the high side of
the uncertainty range reported.
The retrograde solubility observed in the lower Liquid/(Liquid + ZrO2x )
phase boundary was not reproduced in the present work, but given the experimental uncertainties reported [63] we consider that the agreement is satisfactory.
4.2.2.2
Comparison with Thermochemical Data
-5
x 10
Exp. 1100 oC
Exp. 1300 oC
Calc.
(O) Exp. vs. Calc. Error, %
-6
H(O2)
-7
-8
-9
-Zr
-Zr
-10
-11
-12
12
10
8
6
4
2
0
Zr
(HCP)
950
850
-13
0
0.05
0.1
0.15
0.2
0.25
0.3
Temperature, oC 750
(a) Experimental [67, 68] vs. calculated partial molar enthalpies of oxygen,
H O2 , within the bcc, (bcc + hcp) and
bcc phase elds.
650
0.1
0.2
0.3
0.4
X(O)
(b) Absolute percent dierence between

experimental [69] and calculated chemical potential, (o) (Ref: O2,g ), within
the hcp phase.
Figure 4-2: Experiment vs. calculated comparisons for the solid solution phases in the
Zr O system.
Fig. 4-2(a) shows the calculated and experimental [67, 68] partial molar enthalpies
of oxygen ( H(O2 )) within the bcc,(bcc + hcp) and hcp phase elds. In general, good
agreement was obtained, having a maximum error of 12% at low oxygen concentrations. From Fig. 4-2(a) it can also be seen that H(O2 )hcp remains almost constant
66
up to 15 at.%. After this composition, the molar partial enthalpy starts to increase
sharply. This behavior has been observed for the T i O system and is consistent
with the presence of low-temperature ordered structures [68].
Fig. 4-2(b) shows the absolute percent dierence between the experimental [69]
and calculated (O) within the hcp phase eld. Below 30 at% O there is relatively
good agreement between the extensive experimental information and the calculated
properties. However, when the composition approaches the solubility limit, the error increases noticeably. Near this solubility limit, Komarek and Silver [69] observed
that a black layer, probably oxygen-decient ZrO2 was formed at the surface of the
samples. Although they did not detect any weight changes, the presence of this layer
may have hindered the oxidation of the metallic samples and thus aected the nal
compositions measured. Therefore, these experimental data should have a lower statistical weight than the ones obtained at lower oxygen concentrations.
5
-1.6
-1.7
-1.8
oG J/g-atom
f
-1.9
x 10
oG
f,ZrO2-x exp.
oG
f,ZrO2-x calc.
oG
f,ZrO2
1/2(2-x)oGf,ZrO
-2
-2.1
-2.2
Ref:
O2(g)
Zr(BCC)
Zr(Liq)
-2.3
-2.4
-2.5
-2.6
1600
1700
1800
1900
2000
2100
Temperature,oC
2200
2300
2400
Figure 4-3: The standard Gibbs energy of formation of ZrO2x ,0 Gf, ZrO2x , (experimental [70] and calculated) at the lower phase boundary-Liquid/(Liquid + ZrO2x )-per
0
mole of atoms. Comparison with 0 Gf, ZrO2 and 2x Gf, ZrO2 .
2
Fig. 4-3 shows good agreement between the experimental and calculated standard
Gibbs energy of formation for the ZrO2x high-temperature zirconia phase. As
expected, the Gibbs free energy of formation for the stoichiometric phase should be
more negative than that of the sub-stoichiometric phase. The experimental points
were calculated using Ackerman et al. [70] mass spectroscopy measurements, together
with the phase boundary data obtained in Ackerman et al. [63]. The peak observed
in the experimental 0 Gf, ZrO2x curve is due to the retrograde solubility as measured by Ackerman et al. [63]. This was not reproduced using the present model and
therefore in this temperature range, the discrepancies between experiments and cal0
culations are greater. As expected, 0 Gf, ZrO2x is more negative than 2x Gf, ZrO2 .
2
67
This leads to a negative Gibbs energy change for the reaction [70]:
2x
x
ZrO2 (s) + Zr(s)
2
2
4.2.3
ZrO2x (s)
(4.1)
Calculated Zr O phase diagram.

4800
Gas
4300
Temperature, oC
3800
3300
Liquid
2800
2300
BCC
ZrO2-x
1800
ZrO2-x
HCP
1300
800
Zr6O Zr3O
300
0
0.2
Zr2O
0.4
ZrO2
0.6
0.8
1.0
Figure 4-4: Calculated Zr O phase diagram
Fig. 4-4 shows the calculated phase diagram over the entire compositional and
temperature ranges. It can be observed that the model was able to calculate the
equilibrium (CondensedP hases + Gas) phase elds. In general, good agreement was
observed between experimental and calculated phase diagram data. The invariant
points and phase boundaries calculated using the model developed in this work agree
well with the experimental determinations.
4.2.4
Summary
Several sets of thermochemical and phase diagram data (invariant points and phase
boundaries) have been used to develop a model for the T-x phase diagram for the
Zr-O system. The validity of the data used for the optimization was assessed and
it was found that the data were self-consistent. This was later used to assign the
relative statistical weights during the parameter optimization procedure. In general,
good agreement has been observed between the experimental and calculated phase
diagram and thermochemical properties.
The ordered interstitial phases present in the high-oxygen, low-temperature region
of the HCP phase eld have been modelled as three stoichiometric compounds. Although this does not represent the true nature of the order-disorder transformations
68
in interstitial HCP -based solid solutions, we believe that the Gibbs energies obtained
are a good starting point for more complete optimizations.
Except for a few discrepancies at some invariant points, we have been able to obtain a model that is self-consistent and may serve as part of more complete databases
for metallic and oxide systems. The incorporation of a more complete thermodynamic
model into the studying of several technological applications involving the Zr-O system will allow a better understanding of the processes involved.
4.3
Modelling of the Cu T i O System
As noted in Section 4.1, a thermodynamic model of the Cu T i O system is necessary in order to understand the processes of ceramic/metal interfacial formation
involving ceramic oxides and active brazing alloys. Fortunately, thermodynamic assessments of the CuT i [31], CuO [71] and T iO [34] are available and, providing
these assessments are accurate enough, constitute an excellent starting point for the
development of a complete thermodynamic model of the Cu T i O ternary system.
An important characteristic of this system is that at least two ternary compounds
(based on the M6 X family of compounds) play an important role as common reaction
products during ceramic/metal joining operations involving oxide substrates and T ibased active brazing alloys. In this section, a model describing the thermodynamic
behavior of the Cu T i O system over the entire compositional range, as well as
over a limited temperature interval is presented. As will be seen later in this thesis,
the precise calculation of the thermodynamic properties of this system allows the
understanding of reaction layer sequences in a variety of ceramic oxide/metal active
brazing processes.
4.3.1
Experimental Phase Diagram of the Cu T i O System.
Kelkar et al. [72] studied the ternary Cu T i O phase relations at 9450 C around
the T i-rich corner of the phase diagram. In order to do so, they established threephase equilibria through the mixing of powders of T i, T iO2 , Cu and CuO with
carefully controlled proportions to yield the desired overall composition. From their
experiments, Kelkar et al. [72] identied the formation of two distinct M6 X-type
compounds, T i4 Cu2 O and T i3 Cu3 O. It was found that these compounds have independent phase elds at 9450 C. These two compounds belong to the Fd3m space
group and are iso-structural with the carbides [23]. (See Table 4.2):
4.3.2
Thermal Stability range of M6 X compounds
In order to obtain a thermodynamic description of the M6 X compounds as complete

as possible, it is necessary to obtain some experimental information regarding the stability ranges of the compounds in question. Kelkar and Carim [23] prepared samples
of the compounds T i4 CU2 O and T i3 Cu3 O from T i, T iO2 and Cu powders. Using
69
Table 4.2: Phase compositions from experiments by Kelkar et al. [72]
Phase composition (at. %)

Sample
58.5
29.1
T i2 Cu
2.3
65.1
32.6
19.3
79.7
1.0
13.1
57.7
29.2
T i2 Cu
2.5
64.5
33.0
1.7
49.3
49.0
T i4 Cu2 O
12.4
58.5
29.1
T i3 Cu3 O
12.7
47.0
40.3
T i[O, Cu]
31.0
68.1
0.6
T i3 Cu3 O
13.6
43.6
42.8
T iO
53.0
47.0
0.04
Cu[O, T i]
0.0
6.3
93.7
T i3 Cu3 O
16.1
43.7
40.2
T i[O, Cu]
32.9
66.6
0.5
T i3 O2
12.4
T iCu
Cu
T i4 Cu2 O
Ti
T i[O, Cu]
2
T i4 Cu2 O
1
Phase
42.7
57.1
0.3
DTA analysis techniques, the authors were able to determine the temperature range
of stability of both compounds. Both M6 X compounds were observed to be stable at
room temperature. The compounds were then heated under inert atmospheres at a
relatively low heating rate (30 C/min) until melting was detected. The melting points
of T i4 Cu2 O and T i3 Cu3 O were found to be 1127 30 C and 1112 40 C, respectively.
4.3.3
Determination of Thermochemical Properties of M6 XCompounds
In their work, Kelkar et al. [72] used the phase compositions obtained from the microprobe analysis to determine the thermochemical properties (at 9450 C) of the M6 X
compounds T i4 O2 O and T i3 Cu3 O. Unfortunately, their analysis was based on binary
descriptions that are outdated. In this work, the most recent binary thermodynamic
descriptions for the Cu O, T i O, Cu T i [30, 34, 71] are used to calculate the
thermodynamic state of the experimental results obtained by Kelkar et al. [72]:
70
In an isothermal ternary system, a three-phase eld represents a region in compositional space that is invariant with respect to the chemical potential of the elements
within each of the three phases comprising this equilibrium condition. Provided that
the binary phases present in this three-phase eld do not exhibit pronounced ternary
penetration (i.e. the third element is insoluble in the binary phases), knowledge of the
individual compositions of each of the phases already thermodynamically described
can be used to determine the thermodynamic state of the system.
In Table 4.2, the three-phase equilibrium region labelled as Sample 1 shows that
T i4 Cu2 O is in equilibrium with CuT i2 and T i[O, Cu]. Moreover, the compositional analysis shows that O is almost insoluble in CuT i2 (a maximum content of
2.3 at.%), while Cu is insoluble in T i[Cu, O] (a maximum of 1.0 at.%). Therefore,
it can be expected that the thermodynamic state of the system can be determined
from binary descriptions of the T i O and Cu T i systems:
1
-28
0.9
-30
0.8
-32
0.6
log(PO 2 )
a(Ti), ref: BCC
0.7
0.5
0.4
-34
-36
-38
0.3
-40
0.2
-42
0.1
0
0
0.2
x(O)
0.4
0.6
-44
0.2
0.4
x(O)
0.6
Figure 4-5: Calculated T i and O activities in the T i O system at 9450 C
Fig. 4-5 presents the calculated T i and O activities in the T iO system at 9450 C.
According to Table 4.2, the Cu content in the T i[O, Cu] phase is very low. Therefore, one could neglect any ternary thermochemical interaction between Cu, T i and
O, and therefore neglect the inuence of Cu on the chemical potentials of both T i and
O in hcp T i[O, Cu]. Using the phase compositions given in Table 4.2, the activity
of T i, with respect to pure bcc is 0.6567, while the value for log(P O2 ) is 37.7899.
Fig. 4-6 shows the calculated activities (with respect to the stable structures at
those temperatures) of T i and Cu in the Cu T i system at 9450 C. If the phase
compositions presented in Table 4.2 are used, the resulting activity of Cu with respect
to the f cc phase is 0.203.
Through determination of the chemical potentials of all the elements present in the
71
1
0.9
0.8
acr(Cu), ref: FCC
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
x(Ti)
Figure 4-6: Calculated T i and Cu activities in the Cu T i system at 9450 C
three-phase eld identied as Sample 1, the thermodynamic state of the system has
been fully determined and it is therefore possible to obtain the Gibbs free energy for
the reaction:
1
4T i + 2Cu + O2 T i4 Cu2 O
(4.2)
2
Using the standard reference states for all the components in Eq. 4.2, we can obtain
the Gibbs free energy of formation of the compound T i4 Cu2 O:
GT i4 Cu2 O = 4(T i)0 + 2(Cu)0 + (O)0
f
= RT 4 ln(aT i ) + 2 ln(aCu ) +
1
ln(PO2 )
2
(4.3)
Using the values given above, the Gibbs free energy of formation of T i4 Cu2 O at
9450 C is 488, 754 J/mol. Per gram-atom, this value is 69, 822 J/g atom. Compared to the Gibbs free energy of formation of neighboring intermetallics, such as
CuT i2 (with GCuT i2 = 6, 410 J/g atom), it can be seen that this M6 X compound
f
is very stable. The high stability of this compound implies that it would take part
in a great number of equilibria, as will be seen in the phase diagram calculations below.
Once the thermodynamic state of Sample 1 is determined, the results regarding
the Gibbs free energy of formation of the other M6 X compound, T i3 Cu3 O, can be
calculated using data from other three-phase equilibrium phase elds. Sample 3 in
Table 4.2 presents a three-phase equilibrium condition in which T i4 Cu2 O, T i3 Cu3 O
and T i[Cu, O] participate. Since the Gibbs energy of formation of T i4 Cu2 O is
already known from the previous calculation, it is possible to use only the binary
description of the T i[Cu, O] to completely determine the state of the system.
72
Table 4.3: Enthalpies and entropies of formation for the M6 X compounds.
Compound
Hf
Sf
T i4 Cu2 O
631, 814
116.51
T i3 Cu3 O
655, 832
141.67
As can be seen in Table 4.2, Cu is present in this phase in negligible quantities and
therefore it is justied to ignore any thermochemical interactions between Cu and
T i, O. Moreover, since T i presents a large thermochemical interaction with O , this
assumption gains even more weight.
According to the results of the microprobe analysis given in Table 4.2, the composition of the hcp T i[O, Cu] phase in thermodynamic equilibrium with T i4 Cu2 O
and T i3 Cu3 O is 31.0 at.% for O and 68.1 at.% for T i. With these values and aided by
the results presented in Fig. 4-5, the chemical activity of T i is 0.210, while the value
for log(PO2 ) is 35.0198. Using this values, and the reaction
1
3T i4 Cu2 O 2T i3 Cu3 O + 6T i + O2
2
(4.4)
the Gibbs free energy of formation for the compound T i3 Cu3 O is given by:
GT i3 Cu3 O =
f
1
1
3GT i4 Cu2 O RT 6 ln(aT i ) + ln(PO2 )
f
2
2
(4.5)
Using the values obtained for the chemical potentials of T i and O in hcp
T i[Cu, O] and equation Eq. 4.5, the compound has a Gibbs free energy of formation of 483, 261J/mol, or 69, 037J/g atom at 9450 C.
From this analysis, the Gibbs free energies of formation at 9450 C of the ternary
M6 X compounds have been obtained. The result, however, is not complete as no
information regarding the enthalpies and entropies of formation has been obtained.
In order to predict the thermal stability range of the M6 X compounds it is necessary
to use the melting points of the pure compounds obtained by Kelkar and Carim [23].
Using these experimental melting points, together with the thermochemical data at
9450 C it was possible to obtain the values for the enthalpies and entropies of formation of both compounds (See Table 4.3).
Due to its high oxygen anity

It is not clear wether these compounds melt congruently [23]
73
4.3.4
Modelling of the M6 X Compounds and Phase Diagram Calculation for the Cu T i O System
In order to obtain a relatively accurate model of the two M6 X compounds, information regarding the atomic arrangement of the elements within the unit cells must be
used (See Table 4.4):
Table 4.4: Structural information of T i4 Cu2 O and T i3 Cu3 O [73], as presented by Kelkar
and Carim [23]
Compound
T i4 Cu2 O
T i3 Cu3 O
System
Cubic
Cubic
Space Group
Fd3m
Fd3m
Prototype
T i2 N i
T i2 N i
Atoms per unit cell
112
112
48 T i
48 T i
32 Cu
32 Cu
16 O
16 O
16 T i
16 Cu
According to Table 4.4, there are four distinct atomic sites within each unit cell of
the M6 X compounds. A suitable sublattice model for these M6 X compounds would
be (T i)3 (Cu)2 (O)1 (T i)1 for T i4 Cu2 O and (T i)3 (Cu)2 (O)1 (Cu)1 for T i3 Cu3 O. Although in the present work both compounds are modelled as stoichiometric and no
specic sublattice denition is essential for the correct representation of the thermodynamic description of the phases, the present description can be useful when
considering the solubility of other elements in these two compounds [74]. Once the
parameters for the thermodynamic description of T i4 Cu2 O and T i3 Cu3 O phases have
been obtained through optimization, the full ternary phase diagram can be calculated
(See Fig. 4-7).
Fig. 4-7 shows the ternary phase diagram for the CuT iO system at 9450 C. The
triangles with the numbers indicate the experimental data points obtained by Kelkar
et al. [72] and presented in Table 4.2. As can be seen, in general there is good agreement between the experimental and calculated phase equilibria.
4.3.5
Modelling the Solubility of Al in M6 X Compounds in the CuT iO

System.
In ceramic/metal brazing operations involving Al2 O3 and active metal brazes based
on the Ag Cu T i ternary alloy, reaction products belonging to the M6 X-type
74
1.0
0.9
(O
0.8
Mo
le F
rac
tio
nx
TiO2
Ti2O3
TiOx
0.4
0.3
Ti
Ti4Cu2O
0.1
0
bcc0
CuO
Cu2O
5
4
Ti3Cu3O
3
1
Liq
0.2 CuTi2 CuTi
Mole Fraction x(Cu)
1.0
fcc
Figure 4-7: Calculated Cu T i O system at 9450 C.
compounds have been observed [74]. In many of these cases, the reaction products
involved show a tendency to dissolve aluminum, which results from the dissolution
of Al2 O3 into the brazing alloy. In some cases, the reaction products have compositions corresponding to (T i, Al)4 Cu2 O [75]. Under dierent chemical conditions,
compounds with the T i3 (Cu, Al)3 O have also been observed [76]. It is therefore necessary to account for the solubility of Al into the M6 X compounds.
Kelkar and Carim [74] studied the Al solubility in the M6 X compounds T i4 Cu2O
and T i3 Cu3 O by mixing Al, T i and CuO powders and arc melting them under Ar.
The resulting pellets were subsequently annealed at 9450 C for 8 days, to allow complete equilibration. By using XRD and EPMA techniques, they were able to determine solubility limits for both M6 X compounds. While the T i4 Cu2 O compound was
shown to exhibit a limited Al solubility (1.5 at.%), the T i3 Cu3 O compound exhibited
a relatively large solubility range, going from T i3 Cu3 O to T i3 Cu2 AlO. Using the
XRD results, Kelkar and Carim [74] were able to identify the location of Al in the
M6 X structure, which is the location identied as c in Table 4.4. This means that
for the T i3 Cu3 O compound, Cu and Al form a random solution in the c site:
(T i)3 (Cu)2 (O)1 (Al, Cu)1
With the compositions of the samples analyzed by Kelkar and Carim [74] and using
a similar strategy to the one outlined in 4.3.3, it was possible to obtain the Gibbs free
energy of formation for the T i3 Cu2 AlO Al end-member of the (T i)3 (Cu)2 (O)1 (Al, Cu)1
phase at 9450 C (580, 000 J/mol). In order to obtain a temperature dependence for
this Gibbs energy of formation, it was further considered that this compound undergoes melting at the same temperature as the T i3 Cu3 O phase:
75
Table 4.5: Enthalpies and entropies of formation for the M6 X compounds.
Compound
Sf
T i3 Cu2 AlO
4.3.6
Hf
850, 000
221.5
Summary
In the present section, the solid region of the Cu T i O system has been critically
assessed using both thermal and phase diagram information. The thermodynamic
stability of the important M6 X compounds has also been determined. Because of its
usual presence in ceramic/metal joining operations, Al has been incorporated to the
thermodynamic description of the T i3 Cu3 O phase.
As will be seen in Chapter 5, the M6 X compounds of the Cu T i O system are
usually present in ceramic/metal joining applications. Therefore, the thermodynamic
assessment of such compounds is of fundamental importance if one is to understand
the reaction phenomena occurring during ceramic/metal joining.
4.4
Critical assessment of the Cu Zr O liquid phase.
As noted in Section 3.3, in order to understand the nature of the chemical interactions
occurring at the interface between Cu-based brazing alloys and ZrO2 -based ceramic
substrates, an accurate description of the thermochemistry of Cu Zr O melts is
necessary. Fortunately, thermodynamic descriptions for the Cu Zr, Cu O and
Zr O are available in the literature [48, 52, 71]. In the next section it will be seen
that third-order interaction parameters are not likely to play an important role in
determining the chemical behavior of Cu Zr O melts.
4.4.1
Cu Zr O Melts
Despite the importance of its use as a de-oxidizing agent in copper-based melts

(See Haessler et al. [77]), the eect of zirconium on the thermochemistry of oxygencontaining copper melts has been reported only once [78]. In that work, Sudavtsova
[78] used the EMF technique to measure the oxygen chemical potential as a function of
zirconium content in copper melts through measuring the voltage signal across an electrolytic cell using zirconia-based membranes and solid metal-metal oxide powder mixtures as reference electrodes. From these measurements (made in the 1423 1523 K
temperature range), Sudavtsova [78] obtained the following results:
1
i) The slope of a plot of log (aO ) vs. log (% Zr) had a value of 2 , which, according to the reaction
Zr + 2 [O] ZrO2
(4.6)
indicates that as soon as Zr is added to an O-containing Cu melt, ZrO2 precip76
itates, according to the reaction constant, K:

K=
aZrO2
(4.7)
1/2
aO aZ r
ii) By plotting the relationship between log (fO ) vs. % Zr , Sudavtsova [78] found
that the interaction parameter eZr , dened as
O
eZr =
O
log (fO )
%Zr
(4.8)
%Zr0
had a value of 220. In terms of the Raoultian activity coecient , the equivalent interaction parameter , can be calculated using(See Lupis [19, chap.
VII,IX]):
MZr j
MCu MZr
eO +
75, 000
(4.9)
Zr = 230
O
MCu
MCu
It is possible to use the ionic liquid two-sublattice model, as described by Hillert
et al. [55] and modied by Sundman [80]:
Cu+1 , Cu+2 , Zr+4
O2 , V a
(4.10)
where P = 2 yO2 + Q yV a and Q = 4yZr+4 + 2yCu+2 + yCu+1 . By using the thermodynamic descriptions of the Cu Zr, Zr O and Cu O subsystems from [48, 52, 71],
it is possible to obtain a mathematical description analogous to Eq. A.32 (please refer
to Appendix A):
Gionic = yZr+4 yO2 0 GZr+4 :O2 + yCu+2 yO2 0 GCu+2 :O2 + yCu+1 yO2 0 GCu+1 :O2
+QyV a {yZr+4 0 GZr+4 :V a + yCu+2 0 GCu+2 :V a + yCu+1 0 GCu+1 :V a }
+R T
P {yZr+4 ln (yZr+4 ) + yCu+2 ln (yCu+2 ) + yCu+2 ln (yCu+2 )}

+Q {yO2 ln (yO2 ) + yV a ln (yV a )}
+yCu+1 yCu+2 yO2 {LCu+1 ,Cu+2 :O2 } + yCu+1 yO2 yV a {LCu+1 :O2 ,V a }
2
+yZr+4 yO2 yV a {LZr+4 :O2 ,V a } + QyV a yCu+1 yZr+4 {LCu+1 ,Zr+4 :V a }
+yCu+1 yZr+4 yO2 yV a 0 LCu+1 ,Zr+4 :O2 ,V a

(4.11)
Provided that the thermodynamic descriptions available for the subsystems Cu Zr,
Zr O and Cu O are reliable, the only parameter that in theory needs to be
optimized to obtain an accurate description of the Cu Zr O liquid system is
fO is the Henrian activity coecient of oxygen having 1 wt.% as the standard state, and is
dened as
H
f2 = a2 %2 . It is possible to dene the value of f2 at innite dilution, since f2 must become equal
to the slope of the Henrian line. This means that (f2 )%20 = 1 . For a much more comprehensive
explanation, refer to Lupis [79, chap. VII].
77
that corresponding to the last term of Eq. 4.11, namely 0 LCu+1 ,Zr+4 :O2 ,V a . Since
this parameter is multiplied by the site fractions of all the main components in the
(Cu+1 , Cu+2 , Zr+4 )P (O2 , V a)Q system (yCu+1 yZr+4 yO2 yV a ), it is expected that at
very small dilution values of both Zr and O (yZr yO 0), the relative importance
of this parameter on the overall value of the Gibbs free energy (and the resulting
oxygen chemical potential) would be small. If this assertion is true, it would be possible to use only the binary descriptions to obtain an approximate calculation of the
thermochemical properties of Cu O Zr melts at very small contents of both O
and Zr. Since, as noted above, the presence of even very small quantities of Zr leads
to the precipitation of ZrO2 , the calculations of the thermochemical properties of
Cu O Zr melts were done considering that the only thermodynamically stable
phase was the liquid . To make the calculations somewhat compatible with the experimental data measured by Sudavtsova, the Wagner(See [35] and [19]) approximation
was used :
ln O = ln 0 + O xO + Zr xZ r + O x2
O
O
(4.12)
From the model, the following parameters were obtained:
Table 4.6: Parameters for dilute Cu O Zr solutions, using the Wagner formalism [35].
Parameter
A+
B
T
Range, K
ln 0
4.4
8880
T > 1473
O
O
Zr
O
19.4
38390
T > 1473
5.2
167890
T > 1473
At 1473 K, the expression for the activity coecient of oxygen in dilute Cu O

Zr melts is given by:
ln O = 1.6 6.7xO 108xZ r
(4.13)
From Eq. 4.13, it is obvious that there is a dierence of 2 orders of magnitude
between the values for the Zr obtained by Sudavtsova [78] (Zr 75, 000) and the
O
O
value calculated using the thermodynamic optimizations by [48, 52, 71] (Zr 110).
O
In order to reconcile this dierence, it is necessary to analyze the possible sources
for the inconsistencies both within the thermodynamic model itself and within the
experimental procedure used to measure this thermodynamic quantity.
This is equivalent to assuming that there is a relatively large kinetic barrier for the precipitation
of ZrO2 particles in Cu O Zr melts.
Note that a(O) = O xO Ref : O2,g
78
4.4.2
Assessment of the Validity of the Thermodynamic Model for Cu

O Zr melts.
Given the dierences between the experimental and predicted value for Zr , it is esO
sential to consider the eect of third-order parameters (0 LCu+1 ,Zr+4 :O2 ,V a in Eq. 4.11)
on the chemical potential of oxygen in Cu O Zr melts. According to Sundman
[80], 0 LCu+1 ,Zr+4 :O2 ,V a can be given a compositional dependence :
LCu+1 ,Zr+4 :O2 ,V a = 0 LCu+1 ,Zr+4 :O2 ,V a + (yO2 yV a ) 1 LCu+1 ,Zr+4 :O2 ,V a
+ (yCu+1 yZr+4 ) 2 LCu+1 ,Zr+4 :O2 ,V a
(4.14)
At innite dilution (i.e. pure Cu), the eect of the 0 L parameter cannot be decoupled from the 1 L, since (yO2 yV a ) 1. Furthermore, since Zr is only present
in very small amounts, 0 L and 2 L have almost the same eect, given the fact that
(yCu+1 yZr+4 ) 1. Thus, at innite dilution, the compositional dependence of the
LCu+1 ,Zr+4 :O2 ,V a can be neglected.
Assuming that the binary descriptions for the subsystems Zr O, Cu O and
Cu Zr are correct, reduction of the dierence between the experimental results
by Sudavtsova [78] and the models can only be possible by modifying the ternary
interaction represented by 0 L. Fig. 4-8 illustrates the eect of varying the value of the
0
L ternary parameter on the calculated oxygen chemical potential (O) for a specic
oxygen and zirconium concentration. As can be seen, the eects (1 103 J/mol)
are small , even for the greatest value for the 0 L parameter (i.e. 1 106 J/mol).
In the range of values explored, the relative dierence between (O) calculated with
and without 0 L amounted to less than 0.5 %.
Since Zr is a second order parameter (See Eq. 4.12), the eect of the value of
O
the 0 L ternary parameter on the calculated value of Zr is expected to be much more
O
pronounced than that on (O). Fig. 4-9 shows a plot of Zr vs. 0 L. As can be seen in
O
Fig. 4-9 the eect of varying 0 L is signicant. It is therefore reasonable to assume that
by thermodynamically optimizing this parameter, the experimental value obtain by
Sudavtsova [78] can be reproduced. However, when an extrapolation is made on the
plot depicted in Fig. 4-9 to obtain values for Zr that approach the experimental ones
O
( 75, 000), it is found that the necessary values far exceed what can reasonably
be expected (0 L 1 108 J/mol) . From this analysis it can be concluded that it
is highly unlikely that the dierences between the experimental and calculated Zr
O
Note that the Cu+2 species can be ignored as it is not expected to be present in signicant
amounts until very high levels of oxygen are present in the melt, as observed by Assal et al. [81].
At innite dilution, (O) (O)0 L=0 + xZr 0 L
As a manner of comparison, the value for the 0 L ternary interaction parameter for the AgCu
O system reported by Assal et al. [81] is only 1 104 . Although this is clearly a dierent system
and the thermochemical interactions between Ag and O cannot be compared to those between Zr
and O, the fact that relatively small values for 0 L account for ternary experimental results indicates
that, in ternary metallic oxygen-containing melts, binary parameters are dominant.
79
0.8
500
0.6
0.4
-500
0.2
-1000
-1
-0.8
-0.6
-0.4
-0.2
LCu +1 ,Zr+4 :O
0.2
, J/mol
,Va
0.4
0.6
0.8
1
x 10
% Difference from
Difference from
+1
LCu +1 ,Zr+4 :O
+4
Cu ,Zr :O ,Va
,Va
=0
=0, J/mol
1000
Figure 4-8: Eect of the 0 L parameter on the oxygen chemical potential (J/mol) in a copper
solution with xO = 1 106 and xZr = 1 103 at 1473 K
parameters can be explained by third-order interaction parameters for the ionic liquid
two-sublattice model (Cu+1 , Cu+2 , Zr+4 )P (O2 , V a)Q
-20
-40
-60
-80
Zr
-100
-120
-140
-160
-180
-200
1
0.8
0.6
0.4
0.2
LCu +1 ,Zr+4 :O
0.2
2
,Va
0.4
, J/mol
0.6
0.8
1
x 10
Figure 4-9: Eect of the 0 LCu+1 ,Zr+4 :O2 ,V a parameter on the value for Zr in Cu O Zr
O
melts.
4.4.3
Wagner interpolation method.
In his paper on the the chemical behavior of metallic alloys containing trace amounts
of oxygen, Wagner [82], attempted to obtain an interpolation formula to determine
the activity coecient of oxygen in binary metallic mixtures. According to this
analysis, oxygen atoms are considered to be present in liquid metallic melts as quasi80
Model
Wagner
-2
-4
ln(O )
-6
-8
Zr Wagner= -268
O
-10
Zr Model = -108
O
-12
-14
0
0.005
0.01
0.015
0.02
0.025
x(Zr)
0.03
0.035
0.04
0.045
0.05
Figure 4-10: Comparison between calculated oxygen activity coecient in Cu Zr O

metallic liquids using the two-sublattice model and Wagner [82] interpolation at 1473 K
interstitials . Using an analysis analogous to standard central-atom models, Wagner

obtained the following expression:
(1 XB )
(1 XB )
O =
+
1/
1/
Z
O(A)
O(B) Z
(4.15)
where O is the chemical activity coecient, Z is the coordination number of a oxygen quasi-interstitial in its solvation cell, X(B) is the mole fraction of element B and
O(x) is the chemical potential of oxygen at innite dilution in liquid metal X.
Using Z = 6 and the calculated values for the activity coecient of oxygen at
innite dilution in Cu and Zr, it is possible to obtain an approximate value for the
activity coecient of oxygen at innite dilution in Cu Zr metallic mixtures. In
Fig. 4-10, a comparison is made between calculated (using the two-sublattice model)
and interpolated (using Eq. 4.15) oxygen activity coecient at 1473 K. As can be
seen, there is a crude agreement between both calculations, the agreement improving
as xZr decreases. However, it can be seen that both calculations yield similar values
for Zr (Zr agner = 208 vs. Zr odel 108). Note that there is only a dierence of
O
O,W
O,M
a factor of 2 and that both values dier greatly from the experimentally determined
Z
O r 75, 000.
The main argument being the large dierence in atomic radii between oxygen (0.66) and
A
transition metals ( 1.2).
A
Z = 6 corresponds to the coordination number of a quasi-interstitial of the octahedral type in

a closed-packed structure.
81
4.4.4
Summary
Based on this analysis, it is concluded that despite the existence of recent experimental information regarding the thermochemistry of dilute Cu O Zr melts,
it is necessary to generate more experimental information to shed light on the nature of the discrepancies between existing experiments and calculations using CALPHAD models. For the purpose of this work, it will be considered that the model
(Cu+1 , Cu+2 , Zr+4 )P (O2 , V a)Q , without third-order interaction parameters constitutes a reasonable approximation to the description of ternary Cu O Zr melts.
4.5
Thermodynamic Assessment of the T i Zr O System
Many zirconia-metal interfaces of practical relevance involve thermochemical interactions between the ceramic and active metals, such as T i, Zr, Hf , etc. It is therefore
necessary to have a clear understanding of the thermodynamics behavior of such
zirconia-active-metal systems and, since T i is the active metal most commonly encountered in such applications, it has been determined that the assessment of the
T i Zr O system was essential for the understanding of such interfaces.
4.5.1
Phases Present in the T i Zr O System
The most complete collection of experimental work on the T i Zr O system has

been compiled in Ondik and McMurdie [83, pp.11-14, and pp.136-140]. In this
compilation, several phases have been identied. In this section, only the phases
whose models have been modied (with respect to the T i Zr, T i O and Zr O
binary descriptions [34, 49, 52]) or that appear only in the ternary Gibbs triangle are
described:
Gas Phase: In order to calculate the oxygen-rich side of the T i Zr O phase
diagram, the gas phase had to be considered . The ideal gas model was used,
considering the following species:
(O, O2 , O3 , T i, T i2 , T iO, T iO2 Zr, Zr2 , ZrO, ZrO2 )
Liquid Phase: The liquid phase in the T i Zr O system has been observed
to go from either pure metallic T i or Zr [49] to the purely ionic liquid along
the T iO2 ZrO2 pseudo-binary section [84]. As in the Zr O system, assessed
in Section 4.2, it is therefore necessary to use a sublattice model capable of
reproducing this behavior in a continuous manner. The ionic two-sublattice
model for liquids [55, 80] has been found to present an accurate description of
systems of such characteristics. A detailed description of this model is presented
in A.2.4. In the sublattice notation, this model, for the T i Zr O system,
As in Section 4.2, solid-gas equilibrium calculations were necessary to optimize the properties
of the solid oxide phases.
82
can be represented as
T i+2 , T i+3 , T i+4 , Zr+4
O2 , V a
where
P = 2yO2 + QyV a
Q = 2yT i+2 + 3yT i+3 + 4yT i+4 + 4yZr+4
Solid Solutions: As has been noted in Lee [34], Arroyave et al. [52], Waldner
and Eriksson [57], it is well-known that the stable solid phases of zirconium and
titanium (bcc and hcp) dissolve oxygen interstitially into their octahedral interstitial sites. The metallic solid solutions of the T i Zr O can be represented
with the two-sublattice formalism, with one sublattice occupied by zirconium
and/or titanium atoms and the other one occupied by both oxygen interstitials
and vacancies:
(T i, Zr)1 (O, V a)c
where c corresponds to the ratio of interstitial sites to normal sites in each
structure.
For the bcc phase, the stoichiometry ratio c is equal to 3. For the hcp phase,
the stoichiometry ratio used was 0.5.
The ZrO2x Phases: The ZrO2x compound can exist, at 1 atm., in three
dierent structural modications (see Section 4.2). Of these three phases, the
high temperature ones (ZrO2x (tetragonal) and ZrO2x (cubic) ) have been
observed to dissolve T iO2 in signicant quantities [85, 86]. It is therefore necessary to modify the models used in Section 4.2 to account for this fact. Because
of the experimental diculties involving the study of the higher-temperature
phase (ZrO2x ), and according to the T iO2 ZrO2 assessment presented by
Park et al. [84], the only zirconia phase that was modied was ZrO2x , whose
sublattice model is given by:
T i+4 , Zr+4
1
+4
O2 , V a2
As can be seen, the fact that only T i is considered in the model (as opposed
of titanium species with other valences, such as T i+3 ) implies that the ZrO2x
phase behaves as a solid solution of ZrO2x and T iO2 in the tetragonal, HgI2 ,
structure.
The Rutile Phase: The Rutile phase has been previously modelled by Lee
[34] and others with a description identical to the one used to model the hightemperature zirconia phases. This phase has been observed to dissolve signi83
cant amounts of ZrO2 [85] and therefore, it is necessary to describe it as:

T i+4 , Zr+4
O2 , V a2
Ternary Intermetallic Solutions: In their comprehensive review of the experimental information of the T iO2 ZrO2 pseudo binary system, Ondik and
McMurdie [83, 136-140] report that the existence of two ternary compounds,
with nominal compositions ZrT iO4 and ZrT i2 O6 . Although these phases have
been observed to present certain degree of non-stoichiometry, they are considered to be stoichiometric phases in this work, with a sublattice description:
Zr+4
T i+4
O2
where A = 1 B = 4 for ZrT iO4 and A = 2 B = 6 for ZrT i2 O6 .

Omega Phase: According to [83], a ternary phase, identied as (omega) and
with hexagonal structure has been identied at low temperatures in the T i
Zr O system. From experimental data it has been possible to determine that
this phase (with a broad homogeneity region) results from the transformation
of the high-temperature bcc phase, along the T i3 O Zr3 O iso-compositional
line. The bcc transformation occurs at 12000 C. The phase has been
observed to be stable at low temperatures < 4000 C. Because of lack of enough
experimental data the presence of this phase was ignored and instead it was
considered that the bcc phase could take its place in the phase diagram.
4.5.2
Optimization and Comparison to Experimental Data I: 14500 C

Isothermal Section
One of the most complete characterizations of the phase relations and invariant reactions occurring in the T iZr O system was the study performed by Hoch and Dean
[87]. In this work, Hoch and Dean [87] investigated the isothermal section of the
T i Zr O system at 14500 C. After holding the samples at 14500 C for long enough
time to ensure equilibration, the samples were analyzed using XRD, identifying all
the phases present in any given sample.
According to Fig. 4-11(b) and Fig. 4-4 (see Section 4.2) it can be observed that
O tends to stabilize the hcp solid solution. In Fig. 4-11(a), on the other hand, the
mutual chemical interaction between T i Zr tends to stabilize the bcc phase, with
respect to the hcp phase. The isothermal section of the T i Zr O at 14500 C
reported by Hoch and Dean [87] shows that, at least at this temperature, the bcc
1
phases stability range extends from the T iZr binary up to the xO = 3 composition.
In Fig. 4-11(a) it is shown that the maximum stability of the bcc phase is observed at
the xT i = xZr = 0.5 composition. It is therefore reasonable to expect that this will
also be the case for the isothermal section.
As can be seen in Fig. 4-12, the agreement between the experimental [87] and
84
2000
2000
Liquid
1800
BCC
1600
HCP
1200
800
HCP
500
600
0.2
0.4
0.6
0.8
1.0
0.1
0.2
Mole Fraction Zr
0.3
0.4
0.5
Mole Fraction X(O)
(a) Calculated T i Zr phase diagram,

from assessment by Kumar et al. [49].
Ti2O3
1000
0.6
0.7
Ti3O5
Ti4O7
TiO2
1000
1400
TiO
BCC
TiOx
Ti3O2
1500
Temperature, 0C
Temperature, 0C
Liq
(b) Calculated O T i phase diagram,

from assessment by Lee [34].
Figure 4-11: Calculated phase diagrams belonging to the T i Zr O system.

1.0
Gas
0.9
TiO2
Ti4O7
Ti3O5
Ti2O3
Mo
le F
rac
tio
nx
(O
ZrO2
ZrTiO4
0.8
0.6
0.5
TiO
0.4
0.3
hcp
0.2
hcp
0.1
0
1 Phase
2 Phases
3 Phases
bcc
0.2
0.4
0.6
Mole Fraction x(Ti)
0.8
1.0
Figure 4-12: Comparison between calculated and experimental ( [87]) T iZrO isothermal
section at 14500 C.
calculated 14500 C isothermal section of the T i Zr O phase diagram agree well.

Except for very few regions of the section, the single, double and triple phase regions
of the system have been accurately represented. As can be observed, the ZrO2x
phase is in thermodynamic equilibrium with most of the phases present in the system.
85
The reason for this is the large thermodynamic stability of this phase , with respect
to any other phase in the system. The absence of any intermediate suboxide in the
Zr O system is also a consequence of the great stability of this phase (see Fig. 4-4
in Section 4.2). It can also be observed that the maximum oxygen solubility observed
in the ternary bcc phase is about 30 at.%, as has been observed in Hoch and Dean
[87]. This maximum solubility is not an artice of the model (from the description in
4.5.1 it can be observed that the bcc phase could, in principle, dissolve much greater
amounts of oxygen in its relatively open lattice.) and this constitutes additional verication of the assessment, at least for this isothermal section.
4.5.3
Optimization and Comparison to Experimental Data II: Liquidus

Surface
Hoch and Dean [87] also examined the liquidus surface of the T i Zr O system up
to the T iO2 ZrO2 composition. During the optimization of this set of experimental
data it was observed that the liquid description had to be modied extensively, adding
fourth-order interaction parameters,
th
Gion,ex,4 = L0 i+2 ,Zr+4 :O2 ,V a + (yO2 yV a ) L1 i+2 ,Zr+4 :O2 ,V a

T
T
in order to obtain a reasonable liquidus surface, in agreement with the experimental
data. Despite numerous eorts at optimizing the thermodynamic parameters of the
liquid, bcc and hcp phases, it was not possible to obtain complete agreement between
the observed invariant reactions and the calculated ones. Several of these invariant
reactions involved the ZrO2x and ZrO2x phases. Since the thermodynamic optimization of the T iO2 ZrO2 pseudobinary was done independently (based on the
assessment by Park et al. [84]), the number of degrees of freedom available for the
optimization was not the adequate, as the optimization of the T iO2 ZrO2 pseudobinary (see below) was deliberately xed, once a satisfactory description of this
isopleth was obtained.
i.e. Large Gibbs free energy of formation.
86
calc
L bcc
L hcp + T iO + ZrO2
L bcc + hcp + ZrO2

L bcc
L hcp + T iO + ZrO2
N/A
L + bcc hcp + ZrO2
L bcc + hcp + ZrO2
L T iO + T i2 O3 + ZrO2
N/A
L T i2 O3 + T iO2 + T iZrO4
N/A
N/A
N/A
L T iO + ZrO2 + ZrO2
L + hcp bcc + ZrO2
exp
Reaction
L T iO + T i2 O3 + ZrO2
L T i2 O3 + T i3 O5 + ZrO2
L T i2 O3 + T iO2 + T iZrO4
L + T i3 O5 T i4 O7 + ZrO2
L + ZrO2 T i4 O7 + T iZrO4
L + T iO2 ZrO2 + T iZrO4
at.%O
T, 0 C
87
66
65
65
65
58
58
34
54
38
47
19
42
25
34
35
36
30
N/A 20
N/A 31
N/A 31
66
N/A 36
60
N/A 12
49
19
31
20
1750 1665
1930 2012
1825 1798
1845 1850
1650 1714
1700 1635
N/A 1790 N/A
N/A 1710 N/A
N/A 1730 N/A
28
N/A 1650 N/A
32
N/A 1700 N/A
46
16
39
38
exp calc exp calc exp calc
at.%T i
Table 4.7: Comparison between calculated and experimental invariant points in the T i
Zr O system. Experimental temperatures have an estimated variation of 10. Note: N/A
means that the model used is not able to calculate such an invariant reaction.
Table 4.7 shows the calculated and experimental invariant reactions obtained by
Hoch and Dean [87]. One of the most common discrepancies found in the table is the
identity of the ZrO2 phase involved in the invariant reactions. In general, the identity
of the ZrO2 is inverted. For the experimental invariant reactions involving ZrO2 ,
the corresponding calculated ones involve ZrO2 and viceversa. In their experimental
work, Hoch and Dean [87] report XRD analyses for their experiments on the ternary
T i Zr O section at 14500 C. For the liquidus surface determination, they do not
report any XRD measurements so the identity of the phases involved in the detected
invariant reactions is not unambiguous. However, if it is assumed that the phases
taking part in the observed invariant reactions have been positively identied, then
the discrepancies between calculation and experiment can be due to deciencies in
the model.
In their assessment of the T iO2 ZrO2 pseudo-binary section, Park et al. [84]
report that, although there is some evidence regarding the existence of a peritectic
reaction ZrO2 ZrO2 + L at 23000 C reported by Shevchenko et al. [88], there
is a great discrepancy with other experimental results [85, 89], possibly due to the
experimental diculties involved in measuring phase relations at such high temperatures. If the peritectic reaction reported by Shevchenko et al. [88] really exists, and
the T iO2 ZrO2 system were optimized taking this information into account, then
the correct ZrO2 phase could be obtained.
Max/Min/Cong. Melt
Peritectic
Eutectic
1.0
0.9
Mo
le F
rac
tio
n
x(L
iq,
O)
0.8
0.3
0.2
0.1
1973
2675
0.5
0.4
ZrTiO4
0.7
0.6
2049
1830
2496
ZrO2
ZrO2
1720
m
1850 A 1798 B1794
hcp
2130
2012
2015
TiO2
ZrTiO4
1717 1635 1900
1705
1712
1690 1660
1768
1731
Ti2O3
Ti3O5
1714
1825
1770
TiO
1680
1665
1705
1830
1905
hcp
bcc
1807
1710
0
0
0.2
0.4
0.6
Mole Fraction x(Liq,Ti)
0.8
1.0
Figure 4-13: Calculated liquidus surface for the ternary system T i O Zr.
In addition to discrepancies regarding the proper ZrO2 involved in the rst 2

invariant reactions presented in Table 4.7 (L bcc + hcp + ZrO2 ,L bcc +
hcp + ZrO2 ), there is a discrepancy concerning the actual character of the reactions.
Fig. 4-13 shows the calculated liquidus surface for the T i Zr O system. The
eutectic reactions L bcc + hcp + ZrO2 ,L bcc + hcp + ZrO2 are marked with
A and B. From this gure, it can be seen that, in order for these two reactions to
88
exhibit a eutectic character, it would be necessary for the monovariant line involving
the L, ZrO2 andbcc (m) to exhibit a maximum. Using conventional optimization
techniques [90], it is not currently possible to dene the character of an invariant
reaction as experimental information entered into the error minimization procedure.
The character of the invariant reactions was therefore not considered in the model
optimization.
At the very end of Table 4.7, it can be seen that there are three invariant reactions
involving phases close to the T iO2 ZrO2 pseudo-binary that cannot be calculated
with the model parameters obtained with the optimization. In 4.5.4 a more detailed
discussion of the discrepancies in this region of the liquidus surface is presented.
4.5.4
Optimization and Comparison to Experimental Data III: T iO2

ZrO2 pseudobinary section
As mentioned in 4.5.2, [84] have already developed an assessment for the T iO2 ZrO2
pseudobinary section of the T i Zr O system. Because of the pseudobinary character of this vertical section, this assessment can be used, in principle, independently
of the rest of the T i Zr O system. The fact that the T iO2 ZrO2 pseudobinary
is virtually independent of the rest of the phase diagram can be best illustrated by
observing the behavior of the liquid phase . This can be best illustrated by examining
the behavior of the liquid phase, from the metallic region of the compositional space,
to the ionic oxide. Fig. 4-14 shows the site fraction of the dierent T i charged species
(T i+2 , T i+3 , T i+4 ) on the cationic sublattice (see A.2.4) for a constant titanium composition (10 at.%) and varying oxygen mole fraction. In the metallic region of the
system, and up to high oxygen mole fractions (xO 0.6), the only titanium charged
species is T i+2 . This changes abruptly when the oxygen mole fraction reaches the
T iO2 ZrO2 section. At this point, T i+4 is virtually the only titanium species in the
melt. Since in the T iO2 ZrO2 pseudobinary the only metallic cations are T i+4 and
Zr+4 , even the thermodynamic description of the liquid phase is independent of the
rest of the assessment.
As has been noted above, the assessment for the T iO2 ZrO2 system has been
already published [84]. Since Park et al. [84] use their own description of the ZrO2
system (which they do not reference in their work) a consistency problem with the
rest of the database presented in this chapter (specially the one involving the Zr O
binary) became evident as soon as the present model for the Zr O system was
included. Although Park et al. [84] used the same description for the T i O system [34] as the one used in this thesis, it was not possible to reproduce the correct
T iO2 ZrO2 section. The model for the T iO2 ZrO2 system had therefore to be
re-assessed.
Fig. 4-15 shows the calculated T iO2 ZrO2 section with the experimental data
points as compiled by [84]. As can be seen, except for the liquidus region, close to the
Which uses the two-sublattice model for ionic liquids [80].
89
0.35
y +2
Ti
yTi+3
y +4
0.3
Ti
Site Fraction
0.25
0.2
0.15
0.1
0.05
0
0
0.1
0.2
0.3
0.4
0.5
Mole Fraction x(O)
0.6
0.7
Figure 4-14: Site fraction of T i ions in the cationic sublattice of the ionic liquid phase.
xT iO2 = xZrO2 composition, the agreement is good. Phase boundaries for the solid
phases have been reproduced, as well as the temperatures and compositions for the
invariant reactions relevant for this pseudobinary section.
3000
ZrO2
2500
Liquid
ZrO2
1500
ZrTiO4
Temperature 0C
2000
500
ZrTi2O6
1000
ZrO2
TiO2
0.05 0.10 0.15 0.20 0.25 0.30

Mole Fraction x(Ti)
Figure 4-15: Comparison between calculated and experimental T iO2 ZrO2 section. The
experimental data points [85, 86, 88, 89, 9193] were taken from the compilation by Park
et al. [84].
The agreement between experiments and calculations does not rule out possible
deciencies in the model. In 4.5.3, it was shown that there are discrepancies in
the calculated and experimental [87] liquidus surface. Although some of the discrep90
ancies can be attributed to deciencies in the thermodynamic descriptions of the

solid solution phases (bcc, hcp), many are likely to originate from discrepancies in
the T iO2 ZrO2 model. As was said in Section 4.5.3, the fact that the calculated
and experimental invariant points in the metallic region of the system (see rst four
rows of Table 4.7) have the ZrO2 phases inverted could be due to the fact that the
present T iO2 ZrO2 description does not consider the peritectic reaction as reported
by Shevchenko et al. [88]. Because other works contradict this result [85, 92], it was
not possible to assign a signicant statistical weight to this data point.
4.5.5
Summary
Despite the discrepancies observed in the liquidus surface for the T i Zr O system,
it can be considered that the present assessment constitutes a good approximation to
the correct thermodynamic description for the T i Zr O ternary system, over a
wide range of temperatures and composition. In the following chapters of this work,
this description will be particularly useful to rationalize the interfacial interactions
between zirconia-based ceramics and T i.
91
92
Chapter 5
Application of Thermodynamics to
the Study of Ceramic/Metal
Interfaces
5.1
Introduction
In Chapter 3 and Chapter 4, thermodynamic models have been developed to describe several metallic and oxide systems that are of interest in ceramic/metal joining
applications:
i) Using the thermodynamic descriptions of the Ag Cu T i and Cu T i Zr
systems (see Section 3.2 and Section 3.3) it is possible to improve the design
process for active brazing alloys. Important parameters in active brazing alloys,
such as chemical activity of the reactive element and solidus/liquidus points as
a function of alloy composition can be easily calculated for a great number of
alloys without having to perform costly experimental runs.
ii) The thermodynamic description of the Cu T i O system (see Section 4.3)
allows understanding of the reaction sequence observed in a great number of
oxide/metal joining experiments that rely in the addition of T i to otherwise
non-reactive brazes.
iii) The main purpose of this thesis is to provide a methodology for the understanding, analysis and prediction of ceramic/metal interfacial interactions in a general manner. However, to illustrate the methodology, special attention has been
placed to the understanding of the interactions between zirconia-based ceramics
and active metals. Therefore, the thermodynamic models for the T i Zr O
and Cu Zr O (see Section 4.5 and Section 4.4) are essential to this work.
In the present chapter, the CALPHAD (or continuum) thermodynamic models
developed in Chapter 3 and Chapter 4 will be used to analyze the available experimental evidence regarding the interfacial interactions between zirconia and active
metals (principally titanium). By using metastable thermodynamic calculations and
activity diagrams it will be shown that the reaction layer sequences observed under
93
several experimental conditions reported in the literature can be rationalized. Finally,

experimental work on the joining of zirconia-based ceramics and Inconel r superalloys will be presented and the results will be analyzed, using the techniques presented
below.
5.2
Interfacial Interactions between Zirconia and Reactive Metals: Motivation and Fundamentals
As has been noted in Section 4.2, zirconia-based ceramics have several characteristics
that make them attractive for a number of important industrial applications:
i) In Solid Oxide Fuel Cells, SOFC, for example, zirconia-based ceramics are used
because of their ionic transport properties, as well as their high-temperature
resistance [94]. Because of their transport properties, zirconia-based ceramics
are also used as oxygen sensors in a variety of industries and for a variety of
chemical conditions [95].
ii) Zirconia exhibits high strength and fracture toughness, particularly at temperatures below 6000 C [96]. This makes it the material of choice for high-temperature
structural applications [97]. However, because of the great volume change undergone when this material transforms from the tetragonal polymorph to the
monoclinic structure ( V 4 %), the use of pure zirconia is limited when
the ceramic part is to be subjected to thermal cycling conditions. In order to
make zirconia into a practical structural ceramic, dierent aliovalent dopants
(M g, Ca, Y ) have been used to decrease the transformation temperature and
retain either the cubic or tetragonal phases over a greater temperature range .
Provided the detrimental phase transformations are suppressed, zirconia-based
ceramics constitute an excellent choice when a ceramic material is necessary for
high-temperature, high-corrosion applications.
iii) Finally, because of their thermal, oxidation, corrosion and wear resistance, zirconiabased ceramics are the material of choice in thermal barrier coatings for turbine
blades [98].
5.2.1
Motivation for the Study of Zirconia/Metal Interactions
In all the applications listed above, the understanding of the formation and evolution
of ceramic/metal interfaces is key to their development and improvement:
In SOFC applications, the anode material is usually a N i/Y SZ cermet . At the
high operating temperatures of SOFC, the N i in the cermet can act as an electronic
conductor, while the Y SZ is used to reduce the thermal mismatch between the electrode and the electrolyte. Although the use of these composite materials is promising,
See Section 4.2.

Y SZ is zirconia stabilized with yttria.
94
the agglomeration of nickel particles due to sintering is a limiting characteristic in

electrode performance of SOFC [99]. Since sintering is fundamentally a surface energy problem, understanding of the interfacial interactions and wetting characteristics
of Ni/YSZ and Ni-Ti/YSZ systems is essential for the improvement of current SOFC
designs.
As has been noted above, zirconia-based ceramics constitute the materials of choice
for high-temperature structural applications in which the mechanical properties of
metallic parts must be combined with the thermal resistance of ceramics [96]. As
noted in Section 2.6, one of the most common ceramic/metal joining techniques is
active brazing. Therefore, understanding of the thermochemistry of ceramic/active
metal interfacial reactions is of fundamental importance for the design of strong and
reliable ceramic/metal joints.
A limit to the application of zirconia-based ceramics as thermal barrier coatings
for turbine blades is the spalling and de-bonding that occurs as the zirconia/metal
interface is subjected to elevated temperatures for long times. One of the reasons for
this de-bonding is the formation of what is known as a thermally grown layer [100].
This reaction layer is formed by the migration of oxygen, from the zirconia lattice into
the metal and its subsequent reaction with active elements in the metal, such as T i,
Al and Cr. An understanding of active element migration, reaction layer formation
and morphological evolution of such interfaces is fundamental for the improvement
of current thermal coating technologies.
Although this work is mainly focused on application of thermodynamic and kinetic to the study of zirconia/active metal interactions (particularly for C/M joining
applications), the proposed methodologies of analysis can be easily applied to other
cases were zirconia/metal interactions play an important role. It is expected that the
same approach can be applied to any case were interfaces between dissimilar materials
are encountered.
5.2.2
Interactions between Zirconia and Active Elements
One of the most distinctive features of zirconia-based ceramics, is the relatively high
degree of non-stoichiometry that they can withstand without decomposing. This is
particularly true for the cubic phase, which can be reduced from oxygen/metal ratios
of 2.0 to 1.9 (see Section 4.2) . As has been noted above, to suppress thermal shock
and improve the overall toughness of this ceramic material, aliovalent impurities are
added so the cubic phase is stabilized and the detrimental c m or t m transitions,
with their large transformation volume are avoided. The introduction of aliovalent
cations into the cubic lattice introduces vacancies to maintain the electroneutrality
This degree if non-stoichiometry is greatly increased by doping zirconia with aliovalent impuri-
ties.
Aliovalent impurities are those cations whose normal valence state is dierent from that of Zr+4
95
condition. For example, when Y2 O3 is added to ZrO2 , the electroneutrality condition

would be given by [101]:
YZr = 4 [VO ]
According to a thermodynamic model developed for the ZrO2 Y O1.5 system

by Hillert and Sakuma [102] and later rened by Katamura and Sakuma [103], the
presence of oxygen vacancies, due to the introduction of aliovalent impurities, plays
a dominant role in determining the stability of the cubic phase and the lowering of
the c t phase transition. To a rst approximation, the solid solution formed when
aliovalent impurities are added to the zirconia phase can be considered to be ideal.
Besides the extensive oxygen vacancy presence in stabilized zirconia due to the introduction of aliovalent cations, the stabilization of the cubic phase to lower temperatures will also allow the zirconia phase to withstand high degrees of non-stoichiometry,
as is observed in the high-temperature, ZrO2x phase (see Section 4.2). Under localized reducing conditions, as those encountered when zirconia-based ceramics are
put in contact with a highly reactive metal , it is reasonable to expect that oxygen
vacancies can be generated through the partial reduction of the Zr+4 cations.
Perhaps the most readily available evidence of the partial reduction of zirconiabased ceramics is the discoloration (blackening) of cubic zirconia ceramics subjected
to strong chemical [51, 96] or electrochemical [104, 105] reduction conditions. To
understand the partial reduction of zirconia, Janek and Korte [105] proposed several
reactions schemes. In one of them, electrons are trapped at the oxygen vacancies
created by the mass ow of oxygen ions towards the reducing surface of the crystal.
These electrons in turn may be associated with the Zr+4 cations, reducing their overall valence to Zr+3 . The free charge carriers generated by this reduction reaction will
interact with photons, giving the reduced zirconia its characteristic color.
In their analysis of the oxygen-decient region of electrochemically reduced YPSZ, Kanda et al. [104] found, through XPS analysis, that the the reduced regions
within the zirconia crystal exhibited larger peaks corresponding to the Zr+3 valence
sate . Moreover, after analyzing the bonding state of the Y O complexes within
the zirconia lattice, they found that the Y +3 ions did not seem to be aected by the
reduction of the Zr+4 ions, which is consistent with the fact that the Y O bonds
are stronger than the Zr O bonds .
Regardless of the nature of the microscopic reduction conditions, the overall eect
of the localized reduction of the zirconia lattice can be expressed in the following
Such as titanium, zirconium, and so forth.

Kanda et al. [104] also observed Zr+2 and Zr+1 peaks, but they interpreted them as ionelectron defect clusters, rather than an actual reduction of the Zr+4 ion.
The standard heat of formation, per mole of oxygen for ZrO2 is 545 kJ/mol, while that of
Y2 O3 is 586 kJ/mol [104].
96
equation [105]:
X
ZrO2 + 2e + VO ZrO2 + OO
In the early stages of the localized reduction, a moving interface between the reduced and non-reduced regions within the zirconia crystal is expected to appear [105].
The movement of this interface is determined by the ionic and electronic conductivities within the material. Once a steady state is reached, a xed oxygen chemical
potential gradient may be imposed across the zirconia layer due to xed chemical
potentials at both extremes of the ceramic. Although the analysis presented above
was applied to the electrochemical reduction of zirconia through the application of
large electric elds, it is possible to relate these observations to the case in which
the zirconia layer is subjected to an oxygen potential gradient, which is the case in
zirconia/active-metal interactions.
Another important characteristic of zirconia-based ceramics is the dierence in
the diusion rates for the ionic species observed in these materials. In their study of
the diusion mechanisms in Y SZ, Solmon et al. [106] obtained the diusion coecients for the Zr+4 , Y +3 and O2 ions within the zirconia lattice. These results are
presented in Table 5.1.
Table 5.1: Ionic diusion coecients in Y SZ crystals [106]. Note that D = D0 e RT
Ion
D0 ,cm2 /s
Q,kJ/mol
D at
1500 K,cm2 /s
Zr+4
0.093
+461 17
8.21 1018
Y +3
0.248
+462 23
2.37 1017
O2
0.001
+86 2
1.01 106
As can be seen, the cationic diusion coecients are several order of magnitude
lower than the oxygen diusion coecient. The implication of this experimental result is that it is possible to neglect the migration of cationic species towards the
1
ceramic/metal interface, provided no ZrO2 (Zr) + 2 (O) decomposition reaction
occurs.
In the following sections, thermodynamic models will be applied to the analysis of
the ceramic/metal interactions occurring when zirconia substrates are put in contact
with reactive elements. Before presenting the analysis, though, it may be useful to
enumerate the assumptions made in the analysis, as well as their justication:
i) According to the thermodynamic models by Hillert and Sakuma [102] and Kata97
mura and Sakuma [103], the main eect of introducing aliovalent dopants in
zirconia-based ceramics is the increase of the thermodynamic stability of the cubic phase (ZrO2x ) due to the introduction of stabilizing oxygen vacancies. As
a rst approximation, the solid solution between ZrO2 and its dopants can be
consider as ideal. To analyze the thermodynamic interactions between zirconia
solid solutions and active elements, it is possible to consider only the pure cubic
zirconia phase, assuming that is stable (or metastable) at lower temperatures.
ii) Although most of the experimental work on the interactions between zirconiabased ceramics and active metals has been done with doped materials, it is
possible to ignore the interaction between the dopants and the active metal.
This assumption is justied by the ndings by Kanda et al. [104]. As noted
above, in Y SZ, the Y +3 ions did not seem to change their oxidation state even
under highly reducing conditions. This seems to be correlated with the stronger
Y O bonding energy. The same can be expected to happen with zirconia-based
ceramics stabilized with M gO and CaO, which have a more negative free energy
of formation than ZrO2 .
iii) From the diusion experiments performed by Solmon et al. [106], it is evident
that, once subjected to a gradient in oxygen chemical potential, the only ions
that will diuse out of the zirconia lattice and into the zirconia/metal interface
would be the oxygen ions, provided the zirconia phase itself does not decompose.
Contrary to what could be expected with other ceramic systems that do not have
large degrees of non-stoichiometry (i.e. Al2 O3 ), the reaction between zirconia
and the active element would not involve, at least during the rst stages of the
reaction, the full decomposition of the ceramic (ZrO2 (Zr) + 2 (O)) .
5.3
Thermochemistry of Ceramic-Metal Interfaces: Previous Work
Several workers have used thermodynamics to gain a better understanding of the

interfacial interactions between ceramic substrates and liquid (or solid) metals with
active element additions. In this section, the evolution of the thermochemical methods
of analysis of ceramic/metal interfacial interactions will be presented in chronological
order. Although not all the relevant publications have been included in this review,
the works selected represent an accurate picture of the evolution of the understanding of the basic thermochemistry of ceramic/metal reactions. Although considerable
progress has been made in understanding this complex problem, several factors still
limit the applicability of the analyses described in this section. As will be seen below,
the main obstacle to a more generalized understanding of the nature of ceramic/metal
interfaces is the lack of the proper thermodynamic descriptions for the relevant systems under study.
This last assumption is only valid when the zirconia substrate is reacting with titanium in the
solid state. In the liquid state, the dissolution reaction of the substrate may occur more rapidly
than the partial reduction of the zirconia lattice
98
5.3.1
Thermochemistry of Active Metal Brazing Systems
Nicholas [15], examined the eect of adding titanium to several liquid metals, either
pure (Cu T i, N i T i) or alloyed (Cu Sn T i), on the wetting characteristics
of such liquids over alumina substrates. From simple thermodynamic calculations,
Nicholas [15] concluded that in most cases, the activity coecient of T i in the pure
metals was so low that the reaction products resulting from the alumina-liquid interfacial interaction were mostly oxides with O/T i ratios greater than 1, corresponding
to the non-metallic compounds T i3 O5 , T i4 O7 , etc. The contact angle between the
liquid metals and these reaction products was not signicantly dierent than that of
the pure metal and the alumina substrate, which is in accord to the model by Li [9]
described in Section 2.4. Based on this analysis, Nicholas [15] proposed the addition
of solvents that increased the titanium activity (such as Sn and Ag), so wettable
T i O compounds (in general, with a O/T i ration lower than 1) would form at the
ceramic/metal interfaces.
Wang and Lannutti [107] concluded that the sole use of Ellingham diagrams to
determine the reactivity of an active brazing alloy was inadequate. Using Miedemas
semi-empirical model [40], they concluded that elements with a higher oxygen anity than titanium (for example Zr, Hf ) were not necessarily more eective as active
additives to conventional brazing alloys. By comparing partial enthalpies of solution
at innite dilution for several reactive elements (T i, Zr, Cr) with their respective
oxygen anity (measured by the Gibbs free energy of formation of the most stable
oxide for each of the studied active metals), they attempted to predict the reactivity of dierent brazing alloys in contact with several ceramic substrates. A problem
with their procedure, however, was that they did not consider the dissolution of the
ceramic into the liquid. From their analysis, it followed that T i additions to Cu and
even Ag melts would not promote wetting on alumina substrates, since no reaction
was expected. This is clearly contradicted by a wealth of experimental observations
(see Kelkar et al. [72], for example).
5.3.2
Application of Chemical Potential and Phase Diagrams to the

Study of Ceramic/Metal Interactions
A more sophisticated approach to the study of ceramic/metal interfacial reactions

was presented by Inaba and Yokokawa [108]. Through the use of chemical potential
diagrams [109, 110], the authors were able to determine the driving forces for interfacial reactions and possible reaction pathways for a number of dissimilar-materials
diusion couples. By using a chemical potential diagram for the Al T i O system,
Inaba and Yokokawa [108] were able to successfully rationalize the layering of the
reaction products resulting from T i/Al2 O3 couples. Although this approach is generally successful in determining likely reaction paths for dissimilar-materials interfaces,
its application is usually limited to systems with three components as a maximum.
The other limitation to this method is that extensive solubility of the components in
one phase cannot be readily incorporated in the analysis.
99
Over the years, a number of researchers have proposed the use of thermodynamically consistent phase diagrams for the study of ceramic/metal interfacial interactions. Metselaar and Van Loo [111], for example, proposed the use of phase diagrams
to study ceramic-metal interdiusion. By relating the slopes of the tie-lines with the
relative diusion rates of the interdiusing species, they were able to rationalize the
reaction layer sequence and even the morphology observed in dierent metal/carbide
diusion couples. Li et al. [112] performed a critical assessment of the thermodynamic information of the T i Al O system and used the calculated phase diagram,
along with activity maps, to determine thermodynamically possible reaction paths in
T i/Al2 O3 couples. Because of the interstitial nature of solid T i(O) solutions, and the
extremely fast oxygen diusion rates, Li et al. [112] established, as a condition for
a thermodynamically allowed diusion path, the continuous decrease of the oxygen
chemical potential across the dierent reaction layers observed in the ceramic-metal
diusion couples studied.
5.3.3
Diusion Simulations of T i/Al2 O3 Metastable Diusion Couples
A more complete analysis of the T i/Al2 O3 system was proposed by Lee [50]. By
using a one-dimensional simulation of the metastable bcc T i(O, Al)/Al2 O3 diusion
couple, Lee [50] was able to determine the diusion paths along the metastable bcc
T i(O, Al)/Al2 O3 phase boundary for dierent reaction times. Because of the much
higher oxygen diusion coecient within the bcc phase , the bcc T i(O, Al)/Al2 O3
interface was initially enriched with Al, since alumina dissolution into the bcc phase
n(O)
required, due to mass conservation, that the n(Al) ux across the metastable interface
to be constant at all times. By calculating the driving forces for phase precipitation
(see Appendix D ) at very short times after the diusion reaction began, Lee [50] was
able to successfully predict the reaction layer sequence observed in T i/Al2 O3 diusion
couple experiments. The predicted sequences also agreed with the analysis performed
by Li et al. [112].
5.3.4
Use of Activity Diagrams: A More General Approach
Although the methods presented above can be used to describe ceramic/metal reaction experiments in three-component systems, this analysis, becomes almost intractable for higher-order systems. By incorporating more elements to the system
under study, increasingly more information is required regarding the dierent diusion rates and stable (or metastable) phase relations between all the possible reaction
products that could form under dierent thermal and chemical conditions. An example of this limitation can be readily seen in the study of Ag Cu T i/Al2 O3
interfacial reactions. As has been shown in Section 3.2, due to the highly positive
Indicating thermodynamic equilibrium between two phase elds

Oxygen is an interstitial solute in bcc titanium, while aluminum is substitutional.
100
Ag T i interaction, the liquid phase in the Ag Cu T i system presents a miscibility gap in most of the Gibbs compositional triangle at temperatures not much
higher than the eutectic in the Ag Cu system. This phase separation in the liquid,
coupled with the fact that there are ve components in this system, makes it almost
impossible to use the techniques outlined above.
In order to simplify the analysis of the Ag Cu T i/Al2 O3 system, Kelkar et al.
[72] proposed the use of activity diagrams for the Cu T i O system to rationalize
the reaction layers observed during alumina-ceramic active brazing operations. By
using a methodology similar to that outlined in 4.3.3, the thermodynamic properties of the M6 X compounds T i4 Cu2 O and T i3 Cu3 O were coupled to the existing
thermodynamic descriptions for the T iO and CuT i systems to develop thermodynamically consistent two-dimensional activity diagrams for T i and O components in
the CuT iO ternary. By using the criterion of a continuous decrease in the chemical potential of the reacting components Inaba and Yokokawa [108], Kelkar et al. [72]
were able to explain the dierent interfacial layering in several Ag Cu T i/Al2 O3
diusion couples. Because of lack of thermodynamic models for the incorporation
of Al in the M6 X compounds (see 4.3.5), Kelkar et al. [72] did not consider the
stabilizing eect of Al in these compounds and therefore they were unable to predict
the Al content in the T i3 Cu3 O and T i4 Cu2 O layers observed.
5.3.5
Limitations and Application to Zirconia/Metal Interactions
One of the main diculties encountered when trying to understand the nature of
ceramic/metal interfacial interactions lies in the fact that, up until recently, the thermodynamic models describing the systems under study were not available. For example, thermodynamic descriptions for metal metal oxygen systems have been
limited to a few systems (for example, T i Al O [113], Ag Cu O [81]). In the
particular active brazing of ceramic materials, even the widely used AgCuT i alloy
system has not been critically assessed until the present work (see Section 3.2). The
problem becomes even more evident when one considers the available thermodynamic
assessments for the Cu T i O system, which is of fundamental importance if one is
to understand the nature of ceramic oxides-active brazing interfacial interactions, as
most of the reaction products observed in these systems belong to the Cu T i O
ternary .
As has been noted in Section 5.2, there are several industrially important applications where the understanding of zirconia/metal interactions under high-temperature
conditions is essential to the development and reliability assessment of critical components of such technologies. To the best of this authors knowledge, although there
have been some experimental studies on the nature of zirconia/metal interactions,
Kelkar et al. [72] obtained the thermodynamic properties for the M6 X Cu T i O compounds
at one temperature ( 9450 C), and the relative thermodynamic stability of such compounds at
dierent temperatures was not assessed until this work.
101
no thermodynamic analysis of such systems has been developed. In the remaining

sections of this chapter, several examples of zirconia/metal interactions will be investigated and rationalized using the thermodynamic models described in Chapter 3
and Chapter 4. The thermodynamic models developed for the active brazing alloys
will be used to determine the thermochemical conditions at the very early stages
of the zirconia/braze interdiusion processes. By using metastable thermodynamic
calculations, coupled with simple assumptions regarding the likely thermodynamic
behavior of the ceramic/metal couples, it will be shown that a wide variety of experimental conditions can be readily explained. Moreover, by combining the results of
this chapter with the analysis of the metal substrate eects on the thermochemistry
of active metal brazes presented in Chapter 6, a methodology for an improved design
of ceramic/metal joining procedures will be presented.
5.4
Application of Thermodynamic Models to the Study of Zirconia/Ti

Interactions
Over the years, several researchers have performed experiments to understand the
nature of interfacial interactions between zirconia-based ceramics and pure titanium,
both in the liquid and solid states [114117]. Understanding liquid T i/zirconia interactions is important in T i-processing technologies [117] and the development of
metal-matrix composites via the liquid route [116], while the understanding of solid
T i/zirconia interfacial reactions is essential to the development of C/M joining applications [51]. In this section, the thermodynamic model for the T i Zr O system
presented in Section 4.5 is used to rationalize the interfacial reactions occurring in
this system.
5.4.1
Interfacial Reactions between Liquid T i and ZrO2
5.4.1.1
Experimental Evidence on Liquid T i/ZrO2 Interactions
Zhu et al. [116] studied the wettability and reactivity of molten titanium on Y SZ
zirconia substrates. By using the sessile drop experimental technique, they measured
the contact angle between titanium and zirconia at 17000 C under a highly puried
argon atmosphere. The oxygen partial pressure throughout the experiments was kept
below 1022 atm.. From their experiments, Zhu et al. [116] observed the formation
of a thin lm of T iC on the molten titanium surface, due to adsorption of C from
the heating elements. The formation of this lm occurred about 8 min. after the
titanium was melted. The duration of the wetting experiments were later reduced to
8 min.. Zhu et al. [116] used EPMA, AES and EDS to study the interfacial structure
formed due to the interaction between molten titanium and zirconia. The main results are outlined below:
i) A gradual discoloration of the zirconia substrate towards the T i/ZrO2 interface

was observed.
102
ii) The aected region of the zirconia substrate presented evidence of decomposition
of the ceramic into ZrO2 and hcp Zr(O).
iii) Small quantities of T i were observed to be dissolved in the ZrO2 sublattice (less
than 1.5 at.%).
iv) On the titanium side of the interface, a region of hcp T i(O) was observed, with
Zr dissolved substitutionary in the T i matrix.
v) The penetration of Zr within the hcp T i(O) phase was negligible compared to
that of O.
vi) A phase with compositions close to the stoichiometry T i2 ZrO was observed near
the ZrO2 /T i interface.
vii) Through the use of x-ray line scanning, it was determined that the Y2 O3 in
solution with cubic ZrO2 remained virtually inert, with negligible diusion into
the liquid T i.
In similar experiments, Lin and Lin [117] and Lin and Lin [118] studied the
interfacial reactions between molten titanium and CaO-stabilized and Y2 O3 -stabilized
zirconia, respectively. The experiments were performed under a high-purity argon
atmosphere. The titanium/zirconia wetting couple was held at 17500 C for 7 min. and
then cooled at a rate of 40 0 C/min to 01000 C. The specimen was then cooled to room
temperature inside the furnace. The resulting interfacial structure was prepared and
analyzed using an analytical TEM and SEM. The results are summarized as follows:
i) The region within the zirconia substrate in direct contact with T i was observed
to be non-stoichiometric zirconia,(ZrO2x ), with non-stoichiometry coecient
x 0.3.
ii) Secondary hcp Zr(O) spherical particles and ZrO2x nano-crystallites with
x 0.1 within the zirconia substrate were observed.
iii) hcp T i(O, Zr) was observed between the T iO2 and T i layers.
iv) Lin and Lin [117] observed T i2 ZrO precipitates within the saturated hcp
T i(O, Zr) matrix.
v) CaO and Y2 O3 remained in the zirconia layer, without dissolving in the liquid
T i.
vi) Upon cooling, interstitial ordered phases, with stoichiometric compositions T i2 O
and T i3 O precipitated out of the hcp T i(O) layer.
Lin and Lin [119] studied the interfacial reactions between a porous zirconia
mold and a T i 6Al 4V alloy during centrifugal casting. Due to the porosity of
the zirconia substrate, the molten titanium was able to inltrate the mold through
capillary and centrifugal forces. Once a zirconia particle was surrounded by the liquid
103
titanium alloy, it was rapidly dissolved. Oxygen-saturated hcp T i(O) was observed
up to 100 m within the T i alloy.
Lin and Lin [117] reported the presence of small quantities of T iO2 at the calciastabilized zirconia/titanium interface. This was not reported in a later investigation
by Lin and Lin [118] on the yttria-stabilized zirconia/titanium interfacial interactions. Lin and Lin [119] and Zhu et al. [116] do not report the presence of the rutile,
T iO2 phase at the zirconia/metal interface. It is important to note that the three
cases [116, 118, 119] that do not report the presence of T iO2 at the ceramic/metal
interface studied the interaction between yttria-stabilized zirconia and molten titanium. From these experimental results, it may be possible that the aliovalent dopant
in zirconia-based ceramics determines whether T iO2 precipitates at zirconia/liquid
titanium interfaces.
5.4.1.2 Thermodynamic Analysis of Liquid T i/ZrO2 Interactions

In this section it will be shown that by using the thermodynamic models developed
in Section 4.2 and Section 4.5 it is possible to rationalize the experimental results
on the interfacial interactions between liquid titanium and zirconia substrates.
A result common to all reaction experiments presented in 5.4.1 was the fact that
the stabilizing aliovalent dopant (in this case, either Y2 O3 or CaO) did not interact
with liquid titanium. This is compatible with the results described in 5.2.2. Due to
this non-interaction, it is possible to analyze the T i/ZrO2 system independently of
the stabilizing dopant. Thus, the analysis becomes greatly simplied.
Fig. 5-1 presents the calculated T i Zr O isothermal section at 17000 C. Although some of the experiments reported were done at 1, 750C [117, 118], rather than
17000 C [116], it is expected that the results would not be much dierent. The main
features of the isothermal section presented in Fig. 5-1 can be summarized as follows:
i) The liquid phase is stable in two non-connected regions. From the T i O phase
diagram presented in Fig. 7-2, it is evident that the two regions correspond to
melting of the bcc T i and T iO1x phases.
ii) The bcc phase remains stable around the central region of the isothermal section,
at oxygen concentrations below 30 at.%.
iii) The hcp phase is stable in the regions corresponding to high T i and Zr, with a
maximum oxygen solubility of about 30 at.%.
iv) The cubic ZrO2x phase is at equilibrium with both the bcc and hcp solid
solution phases and with the oxides T iZrO4 , T i2 O3 , T i3 O5 and T iO2 .
As can be seen in Fig. 5-1, the 17000 C isothermal section is complex, since a
great number of phases are thermodynamically stable at this temperature. To try to
104
1.0
Mo
le F
rac
tio
n
x(O
ZrO2
0.9
ZrTiO4
0.8
TiO2
Ti4O7
Ti3O5
Ti2O3
0.7
0.6
0.5
0.4
0.3
hcp
0.2
hcp
0.1
0
bcc
0.2
0.4
0.6
0.8
Mole Fraction x(Ti)
liquid
1.0
Figure 5-1: Calculated 17000 C isothermal section of the T i Zr O system.
rationalize plausible diusion paths across the T i/ZrO2 diusion couple, it is possible
to use metastable equilibrium calculations, coupled with simple assumptions regarding the kinetic behavior of the system [120]. Appendix D presents an example of
the methodology developed by Lee et al. [120], applied to the Ag Cu/T i diusion
couple, relevant to many ceramic/metal active brazing applications. In this section,
the same methodology is used to analyze the reaction sequence observed in T i/ZrO2
couples, both in the liquid and the solid states.
At the very early stages of the interfacial interaction between molten titanium and
a solid substrate of the cubic ZrO2x phase, it is expected that a metastable equilibrium will be established between the two phases. Fig. 5-2 presents the metastable
17000 C section of the T i Zr O system, considering liquid and ZrO2x as the
only stable phases. In principle, there are an innite number of thermodynamically
consistent paths that the T i/ZrO2 diusion couple can follow. To nd a plausible diusion path, it is necessary rst to determine the compositions at the interface between the solid ZrO2x and the liquid T i phases. Using the local equilibrium assumption, which is reasonable at this high temperature, it can be postulated that the compositions at the solid/liquid interface will lie somewhere along
the ZrO2x /(ZrO2x + liquid) and the (ZrO2x + liquid)/liquid metastable phase
boundaries.
At these high temperatures, it is reasonable to assume that the rate of solid dissolution into the liquid is much faster than the rate of diusion of the ions within the
ZrO2 phase. Table 5.3 presents the diusion coecients for T i, Zr and O, based
on the diusivities in Table 5.1 and Table 5.2. As can be seen, the rates of diusion within the liquid are much higher than those in all the solid phases. Moreover,
105
1.0
0.9
Mo
le F
rac
tio
n,
x(O
0.8
ZrO2
0.7
0.6
0.5
0.4
0.3
0.2
liquid
0.1
0
0.2
0.4
0.6
0.8
Mole Fraction, x(Ti)
1.0
Figure 5-2: Calculated metastable 17000 C isothermal section of the T i Zr O system.
dissolution of the solid in the liquid implies that the ux of oxygen and zirconium
atoms from the solid to the liquid phase should approximately maintain the stoichiometry corresponding to ZrO2 . The composition of the liquid at the metastable
solid/liquid interface should therefore correspond to the point at which the metastable
(ZrO2x + liquid)/liquid phase boundary is intersected by the line connecting pure
titanium to pure zirconia [120]. This point is indicated by the solid square in Fig. 5-2.
Table 5.2: Diusion coecients in bcc and hcp T i Zr O solutions from [121, chapter
Q
13] and extrapolations. Note that D = D0 e RT .
Diusion
Coecient
bcc
hcp
D0 ,cm2 /s
Q,kJ/mol
D0 , cm2 /s
Q,kJ/mol
Ti
DT i
3.58 104
130, 600
6.6 105
169, 100
Zr
DZr
8.5 105
115, 970
9.713 102
256, 686
T
DOi
8.56 102
138, 200
7.78 101
203, 478
Zr
DO
9.77 101
171, 700
1.32 100
201, 800
Due to the much higher oxygen diusion within the ZrO2x phase, it is expected that an
additional ux of oxygen atoms will come from within the ceramic itself.
106
Table 5.3: Calculated/estimated diusion coecients (cm2 /s) for phases in the T i Zr O
system, at 17000 C. For the liquid phase, the estimation D 1 108 T 3/2 cm2 /s by
Kirkarldy and Young [122, page 93] was used.
element
liquid
Ti
8.8 104
Zr
O
ZrO2
bcc
hcp
N/A
6.5 107
2.2 109
8.8 104
5.8 1014
7.3 108
1.6 108
8.8 104
5.3 106
1.9 105
3.2 106
Once the equilibrium compositions at the metastable solid/liquid interface have

been determined through the procedure outlined above, it is possible to determine
whether another phase can precipitate under these conditions. To establish which
phase precipitates rst, one can calculate the driving force for precipitation for all the
possible phases in the system. Ignoring surface eects, the phase that precipitates rst
is the one with the largest driving force for precipitation [120]. When a metastable
equilibrium between two phases, and is established, this driving force can be
calculated using (see Fig. D-1 in Appendix D):
A
GAa Bb = a (A) + b (B) Gf a Bb
P
(5.1)
where (A) and (B) are the chemical potentials of elements A and B when the
A
phases and are at metastable equilibrium, Gf a Bb is the Gibbs free energy of formation of phase Aa Bb and GAa Bb is the driving force. Only phases with positive
P
driving forces for precipitation will precipitate.
Fig. 5-3 shows the calculated normalized driving forces for precipitation for several phases along the (ZrO2x + liquid) /liquid metastable phase boundary. The
region enclosed with the oval corresponds to the region enclosed within the oval in
Fig. 5-2. As can be seen from Fig. 5-3, there are only two phases (bcc and hcp) that
have positive driving forces for precipitation within the compositional range of the
metastable solid/liquid interface. Of these two phases, hcp always has the greatest
driving force for precipitation, except for a narrow region close to the central region
of the phase diagram. The bcc phase becomes unstable with respect to precipitation
when the oxygen content of the liquid phase is xO 0.1. Note that the driving forces
for precipitation for the T iO and T i2 O3 phases become less negative, as the oxygen
content in the liquid phase increases.
Based on the calculations depicted in Fig. 5-3, it is possible to calculate the
metastable T i Zr O isothermal section at 17000 C, considering only the phases
ZrO2x , hcp and liquid. Fig. 5-4 presents this isothermal section.
Before discussing the possible diusion paths of the system in which the phase
107
Normalized Driving Force for Precipitation
1.0
DGMR(TIO)
DGMR(HCP_A3)
DGMR(BCC_A2)
DGMR(TI2O3)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1.0
0.30
0.35
0.40
0.45
Molar Fraction x(Liq,O)
A Bb
GP a
RT
Figure 5-3: Normalized
0.50
driving forces for precipitation for several phases along
the metastable (solid+liquid)/liquid phase boundary in the metastable 17000 C isothermal

section in the T i Zr O system.
1.0
Mo
le F
rac
tio
n,
x(O
0.9
ZrO2
0.8
Oxygen
Iso-activity Lines
0.7
0.6
0.5
0.4
t2 t1 t0
0.3
0.2
hcp
0.1
0
0.2
0.4
0.6
0.8
1.0
liquid
Mole Fraction, x(Ti)
Figure 5-4: Calculated metastable 17000 C ternary section for the T i Zr O system
considering the ZrO2x , hcp and liquid phases.
sequence is ZrO2 2 x/hcp/liquid, it is worth examining the possibility of any other

further precipitation across the new interfaces created by the formation of the hcp
phase. Fig. 5-5 shows the calculated normalized driving forces for precipitation for
two phases bcc and ZrO2 along the (ZrO2x + hcp)/hcp phase boundary. As can
be seen in this gure, the maximum normalized driving force for precipitation observed at this metastable phase boundary is 0.08, which is considerably less than
108
the driving force for precipitation of the bcc phase at the metastable solid/liquid interface ( Ghcp 0.2). Note that the bcc phase becomes unstable with respect to
P
precipitation as soon as the titanium content in the hcp phase reaches xT i 0.5. The
ZrO2 phase becomes slightly stable with respect to precipitation at titanium contents xT i > 0.5. The possible precipitation of ZrO2 calculated in this gure may be
corroborated by the fact that ZrO2 crystallites have been observed to precipitate at
the zirconia/titanium interface [118], although it is more likely that these crystallites
precipitated from the eutectoid decomposition ZrO2x hcp Zr(O) + ZrO2x
(see below).
Normalized Driving Force for Precipitation
0.10
DGMR(BCC_A2)
DGMR(B_ZRO2)
0.08
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
-0.10
Figure 5-5: Normalized
0.1
0.2
0.3
0.4
Molar Fraction x(hcp,Ti)
A Bb
GP a
RT
0.5
0.6
driving forces for precipitation along the (ZrO2x +
hcp)/hcp phase boundary at 17000 C.
To try to predict the possible diusion path once the interaction between molten
titanium and cubic zirconia ZrO2x has led to the precipitation of an hcpT i(O, Zr)
phase, it is possible to use the metastable ternary section presented in Fig. 5-4. To
simplify the analysis, the composition at the metastable (ZrO2x + hcp) /hcp phase
boundary was considered to be located at the intersection of this phase boundary and
the line connecting the ZrO2 T i initial compositions of the diusion couple.
A likely diusion path at the early stages of the ceramic/metal interfacial reaction is indicated by the thick gray line labelled as t0 . Due to the low diusivity of
zirconium in the hcp phase, it is expected that the hcp T i(O, Zr) layer adjacent
to the zirconia surface will be highly enriched with zirconium, compared to regions
within the hcp T i(O, Zr) phase away from the interface. The ux of oxygen atoms
from the zirconia phase into the interface will be much greater than that of zirconium
due to the great dierence in diusion rates within the oxide (see Table 5.3). Since
cubic zirconia can withstand considerable ranges of non-stoichiometry (see Fig. 4-4),
it is possible for this phase to supply the interface with a continuous ux of oxygen
atoms, without having to maintain the ux ratio between zirconium and oxygen ac109
cording to the stoichiometry of ZrO2 . Because the diusion rate of oxygen within
the hcp phase is at least two orders of magnitude greater than that of zirconium,
it is expected that the diusion path at the early stages of the reaction will follow
the oxygen iso-potentials [50], indicated in Fig. 5-4 as back dashed lines. Regions far
away from the ceramic/metal interface will not have any zirconium, and is expected
that the diusion path in these regions will follow the T i O boundary in the Gibbs
compositional triangle for the T i Zr O system.
As time progresses, considerably higher degrees of non-stoichiometry will be reached
at the zirconia side of the zirconia/titanium interface. The region adjacent to the
1
hcp T i(O, Zr) interface may even decompose trough the reaction ZrO2 Zr + 2 O.
The equilibrium tie-line will then be shifted towards higher oxygen non-stoichiometries
and higher zirconium contents at the (ZrO2x + hcp) /hcp phase boundary. This is
indicated by diusion paths t1 and t2 (thick gray lines in Fig. 5-4), the direction of
the diusion paths is indicated by the dark gray arrow in Fig. 5-4. Because of the
shifting in the tie-lines, the iso-potentials along which oxygen will diuse within the
hcp phase will also be shifted downwards.
Once the diusion couples reacted at high temperatures for 10min, they
were cooled to 10000 C at a rate of about 500 C/hr. On the zirconia side of the
ceramic/metal interface, as temperature decreases, the maximum possible oxygen
non-stoichiometry also decreases until, at 15400 C the highly reduced region within
ZrO2x is expected to undergo the eutectoid transformation ZrO2x ZrO2x +
hcp Zr(O), as shown in Fig. 5-6. According to this transformation, it is expected to
observe precipitation of saturated hcp Zr(O) particles, as well as secondary ZrO2x
crystallites, with lower non-stoichiometries, x, than the parent material. Although
Fig. 5-6 indicates that the cubic phase undergoes the c t m transformation
sequence, the ZrO2x nano-crystallites observed in Lin and Lin [117] and Lin and
Lin [118] preserved the cubic structure due to strain energy eects .
From the results above, it can be seen that metastable thermodynamic calculations for the T i Zr O system can account, at least on a qualitative manner, for
most of the experimental results reported in the literature [116119]. Moreover, the
compositional measurements by Zhu et al. [116] (solid squares in Fig. 5-4) agree well
with the proposed diusion paths. However, the reported presence of T iO2 during
calcia-stabilized zirconia/titanium interactions reported by Lin and Lin [117] cannot
be accounted for in this analysis. Although it is possible that some oxygen could
have been leached out from the zirconia substrate before the whole substrate was dissolved in the molten metal, driving the metastable solid/liquid interface towards lower
contents of zirconium and thus, higher probability for the precipitation of T iO2 , it is
unclear how these T iO2 precipitates could remain stable with respect to re-dissolution
into liquid titanium. The other three experiments reviewed here [116, 118, 119] do
The nano-crystallites did not reach the critical volume to make the c t transformation
thermodynamically possible
110
1800
ZrO2
1750
Temperature 0C
1700
ZrO2+hcp
ZrO2+ ZrO2
1650
1600
1550
hcp+ ZrO2
1500
0.60 0.61 0.62 0.63 0.64 0.65 0.66 0.67
Mole Fraction, x(O)
Figure 5-6: Zr O phase diagram, ZrO2 region.
not report the presence of T iO2 , despite being done at similar temperatures and for
similar times, The only dierence between these experiments and the one reported
by Lin and Lin [117] seems to be the dopant used to stabilize the cubic zirconia
phase. It is unclear, however, how a dopant that remains relatively inert with respect
to reduction (see 5.2.2) could aect whether T iO2 would precipitate or not under
these experimental conditions.
5.4.2
Interfacial Reactions between Solid T i and ZrO2
5.4.2.1
Experimental Evidence on Solid T i/ZrO2 Interactions
In order to study the possible use of partially stabilized zirconia, P SZ, for structural
applications, Suganuma et al. [123] performed solid state diusion bonding experiments between P SZ and steel, using a titanium interlayer. In their experiments,
Suganuma et al. [123] used Y SZ and AISI Type 405 stainless steel and assembled
sandwich structures consisting of steel/titianium/PSZ/titianium/steel. The arrangements were subjected to hot isostatic pressing, HIP, for 15 30 min . at 10000 C,
under a pressure of 100 M P a. After the diusion bond, some of the samples were
analyzed using EPMA, SEM and XRD to determine the interfacial microstructure.
Their ndings can be summarized as follows:
i) From the XRD analysis, it was determined that there were no titanium oxides
present at the ceramic/metal interface.
ii) The EPMA showed that zirconium had not diused into the titanium phase.
111
iii) Oxygen diused into the hcpT i(O) phase, up to a depth of 30 m after 30 min.
at 10000 C.
iv) The zirconia substrate presented the characteristic discoloration (blackening).
In a more recent experiment, Correia et al. [124] studied the interfacial interactions between ytttria-stabilized zirconia and a titanium in its pure state or alloyed
(T i 6Al 4V ). 1 mm thick T i sheets were place between zirconia pieces (with
a thickness of 5 mm) in a sandwich structure. The Y T Z/T i/Y T Z arrangements
were joined at 1162, 1245, 1328 and 14940 C, under an axial pressure of 5 M P a for
15 180 min., under a highly puried argon atmosphere. Their results for the sample
joined at 14940 C, are summarized as follows:
i) The zirconia pieces exhibited extensive blackening.
ii) In the region within the zirconia piece, adjacent to the zirconia/titanium interxO
face, EPMA measurements revealed extensive oxygen depletion, where the xZr
decreased from 2 to 1.6.
iii) The SEM micrograph of this joint exhibited considerable decomposition of the
zirconia substrate.
iv) From the concentration proles, obtained through WDS, it was determined that
there were two distinct layers present on the titanium side of the zirconia/titanium
interface. The rst layer, about 20m thick, presents signicant amounts of zirconium dissolved, with complete oxygen saturation of the hcp T i(O) phase.
On this layer, the xZri) ratio decreased, from 0.5 to 0.
xT
v) From the WDS analysis, it was concluded that, in the region where Zr was
present in signicant amounts, the solubility of O in the metal decreased, compared to the saturation concentration observed in the hcp T i(O) phase.
vi) The second layer presented a uniform oxygen atomic concentration of about
30 at.%.
5.4.2.2
Thermodynamic Analysis of Solid T i/ZrO2 Interactions
As can be seen in Fig. 5-7, the ternary sections for the T i Zr O system at 1000
and 14940 C are very similar. The bcc phase is stable at the central portions of the
metal-rich section of the phase diagram and the zirconia phases (ZrO2 at 1000C and
ZrO2 at 14940 C) are in equilibrium with most of the stable phases at these temperatures. The main dierence in these ternary sections lies on the stability range of the
hcp phase on the titanium and zirconium-rich sides of the phase diagram. As can be
seen in Fig. 5-7, as the temperature decreases, the hcp stability range, especially in
the T i-rich region, narrows. At higher temperatures, however, the T i-rich hcp phase
becomes stable over a wide compositional range. The Zr-rich hcp phase, however,
does not seem to be as stable. The reason for this asymmetry in the stability ranges
for the two hcp phase elds is the asymmetry of the bcc eld itself. As has been noted
112
ZrTiO4
0.6
0.5
rac
0.4
hcp
bcc
0.1
0
0.2
0.4
0.6
0.8
Mole Fraction x(Ti)
TiO2
Ti4O7
Ti3O5
Ti2O3
TiO
ZrO2
0.3
0.2
ZrTiO4
0.8
Mo
le F
Mo
le F
rac
tio
nx
(O
ZrO2
Gas
0.9
TiO2
Ti4O7
Ti3O5
Ti2O3
TiO
0.8
1.0
0.9
Gas
tio
nx
(O
1.0
hcp
1.0
0.5
0.4
0.3
0.2
0.6
hcp
bcc
0.1
0
(a) Calculated T i Zr O ternary section at 10000 C
hcp
0.2
0.4
0.6
0.8
Mole Fraction x(Ti)
1.0
(b) Calculated T i Zr O ternary section at 14940 C
Figure 5-7: Analysis of zirconia/solid titanium interactions. Calculated equilibrium ternary

sections of the T i Zr O system.
in Section 4.5, the bcc phase congruently melts at the Zr-rich side of the metallic
region of the T i Zr O phase diagram.
Despite the dierences between the 10000 C and 14940 C ternary sections of the
T i Zr O system, the solid-state diusion results reported by Suganuma et al.
[123] and Correia et al. [124] could be rationalized using only one of these phase diagrams, taking into account the dierences in diusion rates, as well as the stability
ranges of the bcc phase. Although under these conditions the ZrO2x allotrope of
ZrO2 is the stable phase, the presence of Y2 O3 in the zirconia lattice stabilizes the
cubic form, ZrO2 and allows a much greater degree of non-stoichiometry. In the
following analysis, the eect of aliovalent additions will be included by requiring that
the ZrO2 be stable at these temperatures. In order analyze the diusion path in
the solid T i/ZrO2 system, the procedure described in 5.4.1.2 can be used.
Above 10000 C, the stable phase is in the T i Zr system is bcc . Therefore, at
the very early stages of the inter-diusion process between solid titanium and zirconia, it is expected that a metastable thermodynamic equilibrium will be established
between bcc T i and ZrO2x . Fig. 5-8 presents a metastable thermodynamic calculation for the T iZr O system, at 14940 C, considering only the bcc and ZrO2x
phases. Based on the fact that ZrO2x can withstand considerable ranges of nonstoichiometry before undergoing the ZrO2 (Zr) + 1 (O) decomposition reaction,
2
the initial composition of the zirconia/titanium interface is expected to lie around the
region marked with the black oval in Fig. 5-8. In Fig. 5-8(b), the normalized driving
See Fig. 4-11(a) in Section 3.3.
113
0.8
1.0
0.8
0.7
Mo
le F
rac
0.6
0.5
0.4
0.3
0.2
bcc
0.1
0
0.2
DGMR(HCP_A3)
DGMR(TIO)
DGMR(RUTILE)
0.6
Driving Force for Precipitation
tio
nx
(O
ZrO2
0.9
0.4
0.6
0.8
Mole Fraction x(Ti)
1.0
(a) Metastable 14940 C section of the

T i Zr O system considering only the
ZrO2 and bcc phases.
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
0.1
0.2
0.3
x(bcc,Ti)
0.4
0.5
0.6
(b) Normalized driving forces for precipitation of several phases along the
(ZrO2x + bcc) /bcc phase boundary.
Figure 5-8: Metastable thermodynamic calculation for the T i Zr O system at 14940 C,

considering only the ZrO2x and bcc phases.
forces for precipitation of hcp, T iO and T i2 O3 were calculated along the metastable
(ZrO2x + bcc) /bcc phase boundary. As expected, the hcp phase has a positive driving force for precipitation at both the T i and Zr-rich regions of the phase boundary.
This phase, however, becomes unstable with respect to precipitation at the central
region of the phase diagram, where bcc is the most stable phase(see Fig. 5-7). At the
T i-rich region of the phase boundary, the f cc T iO1x phase starts to become stable,
although its precipitation driving force never exceeds that of hcp. It is expected that
the rst phase to precipitate at the metastable bcc/ZrO2x interface is hcp.
Fig. 5-9(a) presents a qualitative depiction of likely diusion paths in the solid
T i/ZrO2 system. The basic sequences can be summarized as follows:
i) At the early stages of the diusional reaction between ZrO2x and solid bcc
T i(O), it is expected that oxygen saturation of this phase will lead to the precipitation of hcp T i(O), with the same oxygen saturation content. This initial
interfacial composition is indicated in Fig. 5-9(a) as point c.
ii) Due to the fact that ZrO2x can withstand high degrees of non-stoichiometry
without decomposing, the ratio between the oxygen and zirconium uxes across
the ceramic/metal interface do not have to satisfy the stoichiometric condition
ZrO2 .
iii) As oxygen continues to diuse through the hcp T i(O) phase, and into the
bccT i(O) (point a in Fig. 5-9(a)), the bccT i(O) phase will become increasingly
114
1.0
0.6
Oxygen
Iso-activity Lines
0.9
0.8
ZrO2
0.7
tio
n
hcp
0.2
0.2
bcc
a
0.5
0.6
0.7
0.8
Mole Fraction x(Ti)
0.9
1.0
0.6
hcp
0.1
0
(a) Qualitative diusion paths in the

solid T i/ZrO2 system. The ratio between the thicknesses (scaled by the relative molar volumes of the substrates)
of the titanium and zirconium samples
is ZrO2 1.
Ti
0.5
0.4
0.3
0.1
0
0.4
rac
rac
Mo
le F
0.3
Mo
le F
x(O
0.4
tio
n
x(O
ZrO2+hcp 0.5
hcp
bcc
0.2
0.4
0.6
Mole Fraction x(Ti)
a
0.8
1.0
(b) Qualitative diusion path for the

solid T i/ZrO2 system for the case where
the ratio between the thicknesses of
the titanium and zirconia substrates is
ZrO2
T i ..
Figure 5-9: Diusion paths in the solid T i/ZrO2 system. Metastable thermodynamic calculations in the T i Zr O system at 14940 C.
saturated with oxygen. This early stage of the diusion process corresponds to
the experiments reported by Suganuma et al. [123], which observed an oxygen
saturated hcp T i(O) layer directly at the ceramic/metal interface, while the
oxygen penetration into the rest of the titanium substrate only reached 40m.
Because the bcc phase transforms to hcp at temperatures below 10000 C, this
phase was not observed in the study by Suganuma et al. [123].
iv) Once the bcc phase saturates, the hcp/bcc interface will move towards the titanium rich side of the metal substrate, until the remaining bcc phase completely
transforms into hcp. This is indicated by point b in Fig. 5-9(a).
v) Since zirconia can still provide oxygen to the zirconia/metal interface without
having to undergo the ZrO2 (Zr)+(O) decomposition reaction, the zirconium
content on the titanium side of the interface is still zero, as has been observed
by Suganuma et al. [123].
vi) Once the bcc hcp transformation has been completed, hcp T i(O) will rapidly
saturate with oxygen.
vii) If the relative sizes of the zirconia and titanium substrates are such that further
oxygen dissolution into the hcp T i(O) phase involves the decomposition of
ZrO2 , zirconium will start to accumulate at the ceramic/metal interface (point
d in Fig. 5-9(a)). The fact that Zr does not rapidly diuse into the hcp T i(O)
115
phase has to do with its relatively low diusion coecient, compared to that of
oxygen .
viii) Correia et al. [124] observed that this was the case: while the titanium metal
away from the interface seemed to be completely saturated with oxygen (at least
up to a distance of 100m away from the C/M interface), Zr was heavily concentrated in a region within 10m from the C/M interface.
ix) Further decomposition of ZrO2 will eventually lead to a displacement of the
composition of the ceramic/metal interface composition towards the Zr rich side
of the phase diagram, along the (ZrO2x + hcp) /hcp phase boundary. This is
indicated by point e in Fig. 5-9(a).
x) From Fig. 5-9(a) it is observed that further zirconium enrichment of the C/M
interface implies a lower oxygen concentration on the metal side of the interface.
This is also consistent with the experimental results reported by Correia et al.
[124].
xi) A simple calculation considering the initial thicknesses of the titanium and zirconium samples and the moral volumes of the substrates indicates that the nal
equilibrium state of the system corresponds to point e in Fig. 5-9(a).
The diusion path and reaction sequence described above corresponds to the case
where the relative thicknesses of the zirconia and titanium substrates are of the same
order of magnitude, once they are scaled using the dierences in molar volumes of
the substrates. A dierent behavior is to be expected when the ratio of thicknesses
between the zirconia and titanium systems, ZrO2 , approaches innity. Under these
Ti
conditions, it is unlikely that the zirconia substrate will undergo the ZrO2 (Zr) +
(O) decomposition. In this case, no zirconium is expected to be observed on the
titanium side of the C/M interface. Fig. 5-9(b) indicates the expected diusion path
in this case:
i) As before, it is expected that the initial bcc phase will eventually transform to
hcp T i(O).
ii) If there is still oxygen available from the zirconia lattice, complete saturation
of the hcp T i(O) phase is to be expected. This is indicated by point c in
Fig. 5-9(b).
iii) Further dissolution of oxygen into the hcp T i(O) layer will promote the hcp
T iO1x transformation, as indicated in point d in Fig. 5-9(b).
iv) Eventually, it is expected that the titanium layer will further oxidize to oxides
xO
with higher xT i ratios.
As a manner of comparison, Table 5.3 can be examined.
116
An example of this behavior was observed by Arroyave [51], when developing a

ceramic/metal joining technique for zirconia/N i-based super alloys composites with
operating temperatures higher than joints using Cu based brazing alloys. In this
case, a 5m thick titanium layer was sputtered on a zirconia substrate. A ni-based
brazing alloy belonging to the BN i 5 classication was placed between the T icoated zirconia and the N i-super alloy substrates. The formation of a wettable T iO
coating promoted the formation of a sound zirconia/Ni joint. Fig. 5-10 shows an SEM
back-scattered micrograph of this experiment. Using EDS it was determined that the
composition of the layer closely corresponded to T iO1x [51].
TiO1-x
Figure 5-10: Zirconia/N i-based super alloy joint using T i thin lm and N i-based brazing
alloy. Joint created at 11000 C for 15 min.
5.4.3
Comparison with Al2 O3 /T i Interactions
To illustrate the eect that the wide range of non-stoichiometry of ZrO2x has on
the nature of its interfacial reaction with T i, it is worth comparing the diusion paths
described above with that of a very similar and much more studied ceramic/metal system. Fig. 5-11 shows the qualitative diusion paths expected in the Al2 O3 /T i system,
based on the analysis by Lee [50]. Alumina does not withstand its partial reduction,
i.e. the formation of oxygen vacancies, without decomposing. When this ceramic
material is subjected to reducing conditions, it undergoes the Al2 O3 2Al + 3O
decomposition reaction. This is the case when alumina is put in contact with T i at
high temperatures.
In this case, due to mass conservation, the uxes of aluminum and oxygen into
These alloys belong to the N iSiB system and have melting ranges around 11000 C. Since B
is a fast diuser in N i-based alloys, it can diuse out of the braze and into the base metal, increasing
the re-melting temperature of the joint.
117
1.0
0.9
0.8
Al2O3
0.6
0.5
0.4
0.3
0.2
tb
0.2
tb>ta
Ti3Al
bcc
hcp
0.1
b
0.4
ta
a
0.6
TiAl
Mo
le F
rac
tio
nx
(O
0.7
0.8
1.0
Mole Fraction x(Al)
Figure 5-11: Qualitative diusion path in the Al2 O3 /T i system at T 11000 C. Based on
the analysis by Lee [50].
the ceramic/metal interface must maintain the stoichiometric ratio corresponding to

Al2 O3 . Since the diusion rate of oxygen in bcc T i or hcp T i is much higher
than that of aluminum, the metal side of the C/M interface is initially enriched with
Al. By using similar arguments to those used in this section, it is found that, under
this conditions, the T iAl phase has the highest driving force for precipitation and
will be the rst phase to precipitate. This is indicated by point a in Fig. 5-11. As
the reaction progresses, more aluminum will diuse, through the T iAl layer, into
the remaining hcp T i phase. As the interface becomes depleted of Al, the driving force for the precipitation of T i3 Al becomes thermodynamically possible and it
will precipitate between the hcp T i and T iAl layers, as is indicated by point b in
Fig. 5-11. As in the case for ZrO2 /T i interactions, the diusion path will lie parallel
to oxygen iso-activity lines within the hcp T i phase. This is due to the much higher
oxygen diusion rates. As can be seen, in this case no T i O phase is likely to be in
equilibrium with the ceramic oxide. The reason for this is the dierent nature of the
reduction/decomposition process of the ceramic oxide.
5.5
Thermodynamic Study of the Interactions between Zirconia and Ag

Cu T i Brazing Alloys for Ceramic/Metal Joining Applications
As has been mentioned in Section 2.4, due to the dierence in electronic structure
of ceramic substrates and liquid metals, wetting angles in these systems tend to be
greater than 900 . This makes the application of conventional brazing techniques to
C/M joining impossible. For example, Shinozaki et al. [125] studied the contact an118
gle between molten zinc and zirconia and found it to be greater than 1500 . As noted
in 2.3.2 and Section 2.6, the addition of reactive elements, such as T i, Hf , Zr can
promote wetting of an otherwise non-reactive molten metal on a ceramic substrate,
mainly because of the formation of a wettable reaction product. In the particular
case of wettability of molten metals over zirconia substrates, for example, Xue et al.
[126] found that the wetting angle of silver-indium-titanium alloys over zirconia substrates decreased to less than 10 degrees due to the formation of wettable precipitates.
As mentioned in Section 2.6, active brazing is one of the preferred techniques for
joining ceramics to metals. Active brazes based on the Ag Cu T i system are the
alloys of choice when the ceramic in question is an oxide. In this section, the thermodynamic models developed for the Ag Cu T i (see Section 3.2) and Cu T i O
(see Section 4.3) systems are used to understand the available literature regarding
the interfacial reactions observed in zirconia/Ag Cu T i/metal systems.
5.5.1
Thermochemistry of Ag Cu T i Alloys
The thermodynamic model for the Ag Cu T i system was developed in Section 3.2.
As was mentioned in that section, experimental investigations of this system suggest
the presence of an extensive miscibility gap in the central region of the liquidus surface of the ternary. This miscibility gap is an indicator of a strongly positive chemical
interaction between Ag and T i. When liquids belonging to this system are put in
contact with ceramic oxides, the high chemical activity of T i promotes the ceramic
decomposition and the further precipitation of T i O-based reaction products.
Fig. 5-12 shows the calculated chemical activity of titanium as a function of Ag
molar fraction for a Ag Cu T i alloy with xed T i molar fraction xT i = 0.05. As
can be seen, the presence of Ag increases the T i activity to levels up to abcc = 0.55
Ti
when the copper/silver fraction reaches xCu = 0.
xAg
In brazing applications, one of the most important parameters to control is the
temperature at which the brazing procedure takes place. In order to design proper
brazing procedures, knowledge of the solidus and liquidus points is essential. Providing the thermodynamic models are accurate, it is possible to use computational
thermodynamics tools to design brazing alloys suitable for specic applications as
well as their corresponding brazing temperature schedules. As a manner of example, the phase fraction calculation for a commonly used Ag Cu T i brazing alloy
(Cu40Ag 5T i wt.%) is presented in Fig. 5-13. In their study on the active brazing
of zirconia to titanium alloys, Peytour et al. [75] veried the liquidus point for the
Cu 40Ag 5T i wt.% active brazing alloy to be about 8500 C, which agrees well with
the phase fraction calculation presented in Fig. 5-13.
As has been noted in Section 2.3.2, the addition of reactive elements to otherwise
non-reactive liquid metals can promote the decomposition of a ceramic substrate
119
0.55
ACR(Ti) Ref. State: BCC-Ti
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
0.2
0.4
0.6
X(AG)
0.8
1.0
Figure 5-12: Chemical activity of T i in an AgCuT i alloy at 11000 C with the composition
constraint xAg + xCu = 0.95.
1.0
0.9
Liquid
fcc#1
fcc#2
Cu4Ti
0.8
Phase Fraction
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
750
800
850
Temperature, T0C
900
Figure 5-13: Phase fraction calculation for Cu 40Ag 5T i wt.% active brazing alloy.
in contact with it, provided there is a highly negative interaction energy between
the active element and the anion in the ceramic. To illustrate the increase in reactivity of brazing alloys through the addition of titanium, the equilibrium between
ZrO2x and Ag Cu alloys was calculated, considering the dissolution reaction
120
ZrO2 (Zr) + 2(O). Fig. 5-14 presents equilibrium calculations of the interaction
between zirconia and silver-copper brazing alloys. In Fig. 5-14(a) the oxygen dissolution in a Cu 40Ag wt.% braze is shown as a function of titanium content in the
braze. As can be seen, the degree of dissolution increases as T i increases, due to the
strongly negative T i O interaction. As a manner of comparison, the same calculation was performed for a brazing alloy with no titanium but with varying amounts of
copper and silver.
-9
-8
-10
-9
-11
-12
LOG { x(O) }
LOG { x(O) }
-10
-11
-12
-13
-14
-15
-16
-13
-17
-14
-15
-18
-19
0
0.02 0.04
0.06 0.08 0.10 0.12 0.14

W(Ti)
(a) Oxygen dissolution in liquid Ag

CuT i brazing alloy when it reacts with
ZrO2 as a function of titanium content.
Base alloy was Cu 40Ag wt.%.
0.2
0.4
0.6
W(Cu)
0.8
1.0
(b) Oxygen dissolution in liquid Ag Cu

brazing alloy when it reacts with ZrO2
as a function of copper content.
Figure 5-14: ZrO2 (Zr) + 2(O) dissolution in Ag Cu T i brazing alloys at 11000 C.

The thermodynamic models developed in this work were coupled with the thermodynamic
assessment of the Ag Cu O system by Assal et al. [81].
5.5.2
Interfacial Reactions Occurring during Ag Cu T i Brazing of

Zirconia
Peytour et al. [127] and Peytour et al. [75] examined the use of Ag Cu T i brazing
alloys in the brazing of zirconia-based ceramics (Y SZ) to titanium alloys (T i 6Al
4V ). A sandwich structure Y SZ/Ag Cu T i/T i was assembled and brazed in a
vacuum furnace (1 105 torr) at 8700 C for 5 min. and cooled to room temperature
in 34 h.. The brazed alloy used during these experiments had the composition Cu
40Ag 5T i wt.%. The samples were then cross-sectioned and analyzed using EPMA,
TEM and XRD. The results from their experiments are summarized as follows:
i) The region of the zirconia piece adjacent to the ceramic/metal interface presented
signicant discoloration, indicating migration of oxygen atoms from the zirconia
121
lattice towards the interface.

ii) Zirconium did not diuse through the interface, therefore, no decomposition of
zirconia took place during the brazing operation.
iii) A 2 4m reaction layer was observed at the zirconia/braze interface. Through
TEM and EPMA analyses, it was concluded that this phase was an M6 x compound with the stoichiometric composition corresponding to T i4 Cu2 O.
iv) Small amounts of aluminum from the titanium alloy were observed dissolved in
this phase (xAl < 2 at.%).
v) The sequence Ag Cu T i/ (Cu, Al)4 T i/Cu4 T i3 /CuT i/CuT i2 /hcp T i was
observed on the metal side of the interface. This sequence correlates well with
the equilibrium Cu T i phase diagram (see Fig. 3-3(b)).
Yamazaki and Suzumura [128] studied the segregation and reaction layer formation phenomena when zirconia-based ceramics are brazed to themselves using
Ag Cu T i brazing alloys. In their experiments, Yamazaki and Suzumura [128]
used a brazing ller metal based on the eutectic of the Ag Cu system with a composition Ag 26.7Cu 4.5T i wt.%. The joints were brazed in a vacuum furnace
with a pressure P < 6 102 P a. The sample was heated to 8160 C, held at this
temperature for 5 min. and then cooled to 3960 C in less than 5 min .. The samples
were then cross-sectioned and analyzed using SEM and EDX analysis. From their
analysis, it was found that there were two reaction layers adjacent to the zirconia
surface. The rst one had a composition close to the T iO phase. The next layer
was a Cu T i O compound, with a copper-titanium ratio close to xCu 1. The
xT i
exact composition of this interface could not be determined.
Emiliano et al. [115] studied the brazing of zirconia, T ZP , to T i and to itself
using a Ag CuT i brazing alloy with the nominal composition Ag 35Cu1.65T i.
The samples were heated to 9000 C for up to 30 min. The resulting joints were crosssectioned and analyzed using conventional analytical techniques (EPMA, EDS, SEM ).
The results from this study can be summarized as follows:
i) In all cases, the zirconia substrate exhibited signicant discoloration close to the
ceramic/metal interface.
ii) In the case of T ZP/T ZP joints, it was observed that the only reaction product
formed at the zirconia/braze interface corresponded to T iO.
iii) The T ZP/T i joints exhibited a more complex microstructure. In this case, the
rst reaction layer observed at the zirconia/braze interface was T i2 O, followed
by a sequence of complex Cu T i O intermetallic compounds belonging to the
stoichiometries T i3 Cu3 O and T i4 Cu2 O.
iv) On the metallic side of the interface, the phase sequence observed was Ag Cu
T i/Cu4T i/Cu4T i3 /CuT i/CuT i2 /T i.
122
Chuang et al. [129] studied the interfacial phenomena during active metal brazing of P SZ to itself using a Ag 27Cu 3T i wt.% alloy. The joints were brazed
at 9000 C for 10 min. under a vacuum atmosphere of 104 torr. Their results were
similar to those of Emiliano et al. [115] and the only reaction layer observed at the
ceramic/braze interface was T iO.
5.5.3
Thermodynamic Analysis of ZrO2 /Ag Cu T i/T i Brazed Joints
As has been seen above, the nature of the interfacial reactions at zirconia/metal interfaces become increasingly complicated when the number of components in the molten
alloy increases. In principle, a diusion path analysis similar to the one presented in
Section 5.4 may be possible, providing there is a precise knowledge of the thermochemical conditions of the braze alloy and the relative diusivities of all the elements
involved in such reactions are known, at least qualitatively. In the particular case
of the interfacial interactions occurring in the ZrO2 /Ag Cu T i/T i system, very
little is known about the properties of the observed reaction products, especially with
regard to the M6 X compounds. Despite the lack of knowledge of the properties of
this system, it is still possible to examine and rationalize the interfacial reactions
observed, provided an accurate thermodynamic description of the system is available.
Kelkar et al. [72] estimated the thermodynamic properties of the Cu T i O
system at 9450 C based on the phase diagram determination of the ternary section,
coupled with simple thermodynamic calculations considering the phases observed to
be at equilibrium with the M6 X compounds T i4 Cu2 O and T i3 Cu3 O. From this
analysis, they calculated the ternary activity map (described below) for this system.
Assuming that at these high-temperatures local equilibrium could be achieved and
that the reaction path observed in the Al2 O3 /Ag Cu T i/T i system should be
such that the chemical activities of the reacting components decreased monotonically, Kelkar et al. [72] were able to rationalize the reaction layer sequences observed
in alumina/titanium active brazing experiments. The applicability of the analysis by
Kelkar et al. [72] is limited due to the fact that the analysis of the Cu T i O system
was performed at only one temperature.
In Section 4.3, the CuT iO system was thermodynamically assessed taking into
account the most recent thermodynamic descriptions for the CuT i [30], CuO [71]
and T i O [34] systems, the experimental results on the thermal stability limits of
the M6 X compounds [23], and the phase diagram determination of the Cu T i O
system [72]. By using the model for the Cu T i O system developed in this work,
it is possible to rationalize the interfacial interactions between zirconia and active
brazing alloys at arbitrary brazing temperatures.
Fig. 5-15 presents the calculated activity diagram for the Cu T i O system
at 9000 C. In a ternary activity diagram, the phase-state of the system is determined
by the activity of one element, plus the relative compositions of the remaining two
123
bcc
hcp+CuTi2
bcc+hcp
0.9
acr(Ti) Ref. State: bcc-Ti
0.8
0.7
CuTi2+Ti4Cu2O
0.6
0.5
0.4
Ti4Cu2O+hcp
CuTi+Ti4Cu2O
Cu4Ti3+Ti3Cu3O
CuTi+Ti3Cu3O
Ti3Cu3O+Ti4Cu2O
0.3
0.2
0.1
0
0
Liq+Ti3Cu3O
Ti3Cu3O+hcp
Ti3Cu3O+TiO
fcc+Ti3Cu3O
0.1
0.2
0.3
0.4
0.5
fcc+TiO
0.6
0.7
0.8
0.9
x(O)/{x(O)+x(Cu)}
Figure 5-15: Activity diagram for the Cu T i O system at 9000 C
components of the system. In this diagram, horizontal lines represent the activity at
which three phases are at equilibrium.
In principle, the activity of an element should show a monotonic variation across
a thermodynamically stable reaction zone . Based on this constraint, it is possible to
restrict the possible diusion paths and reaction layer sequences in any given diusion
couple. The monotonic variation of the activity of a reacting element implies that
the reaction path observed in an activity diagram should always follow a direction
towards decreasing chemical activity.
The reaction sequence observed by Peytour et al. [75, 127] is depicted in Fig. 5-15
as the dashed gray line with triangles as the markers. From this diagram, it can be
seen that the titanium substrate acts as a continuous source of titanium, maintaining always the saturation composition in the liquid. The titanium activity decreases
monotonically towards the reaction product with the lowest T i activity, T i4 Cu2 O.
Even though the samples were held at the brazing temperature for only 5 min., the
reaction products observed satisfy the local equilibrium condition.
The reaction sequence observed by Peytour et al. [75, 127], however, does not
contemplate the formation of T i3 Cu3 O and T i2 O at the zirconia/braze interface, as
was observed by Emiliano et al. [115]. This latter reaction sequence is also depicted
This requirement, however, does not hold when this element is forced to diuse uphill due to
the diusion of other elements [72]
124
by the dashed gray line with triangles, plus the dashed line with squares. It is important to note that in the brazing experiments by Emiliano et al. [115], the samples
were held at the brazing temperature for 30 min.. The formation of the T i2 O and
T i3 Cu3 O phases seems to be limited by kinetic (i.e. nucleation) factors.
It is important to note that the Cu4T i reaction product observed in Peytour et al.
[75], Emiliano et al. [115], Peytour et al. [127] is not stable at 9000 C. Its formation,
therefore, must have occurred upon cooling of the sample.
5.5.4
Thermodynamic Analysis of ZrO2 /AgCuT i/ZrO2 Brazed Joints
Contrary to the experiments on ZrO2 /T i brazing, when zirconia was brazed to itself
the only reaction products observed where T i3 Cu3 O and T iO [115, 128, 129]. The
reason for this is that the maximum T i activity in the system corresponds to the initial thermochemical condition of the brazing alloy, since there is no external titanium
source. Table 5.4 shows the calculated initial activities for the three dierent brazing
alloys used by Emiliano et al. [115], Yamazaki and Suzumura [128], Chuang et al.
[129]. Comparing these T i activities to the activity map in Fig. 5-15, it is evident
that the maximum activity of these alloys lie below the minimum stability range of
the T i4 Cu2 O. This reaction layer is not expected to form at the zirconia/braze interface during Zirconia/Zirconia active brazing.
Table 5.4: Titanium activities of Ag Cu T i Brazing Alloys.
Alloy
Composition
Brazing
Temperature
Ti Activity. Ref:
Ti-bcc
Ag 26.7Cu
4.5T i wt.% [128]
8160 C
0.20
Ag 35Cu
1.65T i wt.% [115]
9000 C
0.15
Ag 27Cu
3T i wt.% [129]
9000 C
0.19
At the early stages of the zirconia/braze interaction, oxygen will diuse out of the
zirconia lattice. Due to the strong T i O interaction, T i will segregate to the zirconia/braze interface [5, 72]. Since in all these brazing experiments the braze interlayer
was a thin foil of about 100m, it is expected that this titanium segregation will
cause the titanium content in the brazing alloy to decrease as the interfacial reaction
progresses. Yamazaki and Suzumura [128], Chuang et al. [129] report that, according
to EDS measurements, the active braze interlayer was eectively depleted of titanium
after the brazing process was completed. Since most of the titanium will be present in
the reaction layers at the zirconia/braze interface, a diusion path in this case will be125
gin with the highest activity of titanium in the reaction layer immediately adjacent to
the zirconia surface. This diusion path is indicated in Fig. 5-15 as a solid gray line,
beginning in the T iO phase eld, through the T i3 Cu3 O phase, towards the f cc phase.
In order to determine the rst phase to precipitate under these experimental conditions, a metastable thermodynamic equilibrium calculation between ZrO2x and
three Ag Cu T i liquid alloys with compositions indicated in Table 5.4 was performed. From these calculations (see Table 5.5), it was evident that both T iO and
T i3 Cu3 O will precipitate, provided there was enough time for the phases to nucleate.
In all cases, the phase with the highest driving force for precipitation was T iO and
it is therefore expected that this phase will be found adjacent to the zirconia surface.
Contrary to what would be expected, however, there is only one instance [128] in
which T i3 Cu3 O is also present in the reaction zone close to the zirconia/braze interface. The reason for this remains unclear, although it could be related to a lower
stability of the T i3 Cu3 O phase as the temperature increases. The melting point of
this compound was found by Kelkar and Carim [23] to be about 11200 C.
Table 5.5: Normalized driving forces for precipitation at a metastable ZrO2 /Ag Cu T i
interface.
Alloy
Composition
Driving Force
Ti3 Cu3 O
Ag 26.7Cu
4.5T i wt.% [128]
2.81
1.6
Ag 35Cu
1.65T i wt.% [115]
2.54
1.28
Ag 27Cu
3T i wt.% [129]
5.6
Driving Force
TiO
2.73
1.49
Interfacial Reactions in the Zirconia/Cu-Ti-Al/Ni System
Although conventional active brazing alloys based on the Ag Cu T i have been

proven eective in zirconia/metal applications, their applicability is somewhat limited due to their relatively low melting range (800 8500 C). In many technologies
were zirconia-based ceramics play an important role, it is necessary to develop ceramic/metal joining techniques capable of producing joints with higher operating
temperatures. To this eect, brazing alloys based on the Cu T i system were examined.
Partially stabilized zirconia and Inconel 718 r discs ( 10 mm diameter5 mm height
126
were cut and polished to a 1 m surface nish. T i, Cu and Al powders were premixed
and then ball milled for 2 hours to achieve uniform particle size. The resulting mixtures were cold-pressed in a die to make preforms with dimensions 10 mmdia1 mmh.
The zirconia, nickel and brazing preforms were assembled into a sandwich structure
and placed inside a vacuum furnace (see Appendix F). A constant axial load was
applied to the samples .
5.6.1
Experimental Results
In Arroyave [51], a complete description of the active brazing experiments using several braze compositions and temperatures is presented. For the purpose of this work,
and to illustrate the use of thermodynamic tools to rationalize just some selected
experimental results are presented. Note that all the brazing experiments presented
here had holding times of about 1 hr at the brazing temperature.
4
3
Zirconia
Figure 5-16: Zirconia/Metal brazing using Cu 10 wt.% at 10250 C.
Fig. 5-16 shows the zirconia/braze interface of a zirconia/Inconel r joint using a

Cu 10 wt.% brazed at 10250 C. This interface is representative of all the other joints
using this composition and brazing temperatures within the 1025 10500 C range.
The reaction product adjacent to the zirconia surface was found to have a composition corresponding to T i2 O3 . At these high temperatures, signicant dissolution of
N i into the liquid braze was observed (point 3 in Fig. 5-16).
In order to reduce the N i dissolution and to develop brazing alloys with a higher
solidus-liquidus range, brazing alloys with a composition Cu 5 wt.% were used in
zirconia/Inconel r joining experiments. Fig. 5-17 shows two micrographs corresponding to joints brazed at 9500 C and 10250 C. Note that the reaction products at
For details regarding the experiment, refer to Arroyave [51]
127
Table 5.6: Compositions of points indicated in Fig. 5-16, at.%
Cu
2.2
21
14.6
97
Ni
18
0.6
Ti
37
79
59
1.2
Cr
3.2
0.4
Fe
3.2
0.3
60.8
the zirconia/braze interface have dierent stoichiometry. At 9500 C the T i O oxide

can be identied as T iO, while the reaction layer at the higher temperature has a
composition close to that of T i2 O3 . In the joint at 10250 C incipient dissolution of the
nickel super alloy can be observed (point 5 in Fig. 5-17(b)).
Zirconia
1
4
3
(a) Zirconia/Inconel r
5 wt.% Braze at 9500 C.
Zirconia
joint.
(b) Zirconia/Inconel r joint. Cu5 wt.%

Braze at 10250 C.
Cu
Figure 5-17: Brazing experiments using Cu 5 wt.% brazing alloys.
One of the main limitations of active metal brazing alloys based on the AgCuT i
system is their poor oxidation resistance [130]. In order to increase the oxidation resistance of active brazing alloys, Kapoor and Eagar [131] proposed the addition of
aluminum to Cu T i and Sn T i conventional active brazing alloys. In their experiments, Kapoor and Eagar [131] determined that Al eectively reduced the oxidation
rates of these active alloy. Through wetting experiments, it was also shown that the
addition of Al did not aect the reactivity and wettability of the brazing alloys when
they were put in contact with a Al2 O3 substrate. Based on these results, an investigation on the use of Cu T i Al alloys for zirconia/metal brazing applications was
128
Cu
96
96.6 8.6
Ni
1.0
6.9
38
1.6
57.4
Ti
46.8 3.8
Cr
0.4
6.9
Fe
0.3
7.8
53.3
62
performed.
2
1
Zirconia
Figure 5-18: Zirconia/Inconel r joint. Cu 10T i 5Al wt.% braze at 9750 C.
Fig. 5-18 shows a micrograph of the zirconia/braze interface resulting when the
ceramic/metal joint is created using a Cu 10T i 5Al wt.% brazing alloy. According
to Table 5.8, the composition of the reaction layer adjacent to the zirconia substrate
has a composition corresponding to T i3 (Cu, Al)3 O. Extensive dissolution can be observed (point 3 in Fig. 5-18). Fig. 5-19 shows a higher magnication micrograph of
the interface shown in Fig. 5-18. From this gure, it is evident that a third phase
has precipitated at the ZrO2 /T i3 Cu3 O interface. Although the analytical techniques
used in this work do not allow the identication of these small precipitates, it may
be possible that they are T iO particles.
129
Cu
36.8
93.7
30.1
Al
7.4
3.5
8.2
Ni
2.7
0.2
15.1
Ti
38.2
2.7
43.8
Cr
Fe
1.8
14.9
Ti3Cu3O
TiO?
Figure 5-19: High magnication of zirconia/Cu T i Al interface.
5.6.2
Analysis of Experimental Results
The results presented above can be summarized as follows:

i) When the Cu 10 wt.% brazing alloy is used at high temperatures, extensive
dissolution of the nickel based substrate occurs. Under these conditions, the
reaction product observed is T i2 O3
ii) In the case of the Cu 5 wt.% brazing alloy, at the lower brazing temperature,
9500 C, T iO is observed to form at the zirconia/braze interface. When the brazing
temperature is 10250 C, incipient dissolution of the metal substrate leads to T i
scavenging and the resulting reaction product is T i2 O3 .
iii) When the Cu 10T i 5Al brazing alloy is used, although signicant titanium
scavenging occurs, the reaction product at the zirconia/braze interface seems to
130
be T i3 (Cu, Al)3 O.
0.16
acr(Ti) Ref. State: bcc-Ti
0.14
Cu-10Ti
0.12
0.10
0.08
0.06
0.04
0.02
850
Cu-5Ti
900
950
1000
1050
Temperature, 0C
1100
Figure 5-20: T i activities of the Cu 10 and Cu 5 brazing alloys as a function of temperature.
Fig. 5-20 shows the T i chemical activity for the Cu 10 wt.% and Cu 5 wt.%
brazing alloys as a function of temperature. As can be seen, when the brazing alloy
lies within the f cc+liquid two phase eld, the chemical activity of titanium increases.
At higher temperatures, the chemical activity is about 0.03 for the Cu 5 wt.% alloy
and 0.06 for the Cu 10 wt.% braze.
By examining the activity diagram for the Cu T i O system at 1000C in
Fig. 5-21, and using the chemical activities for the two brazing alloys in Fig. 5-20,
it is evident that both T i3 Cu3 O and T iO should be observed at zirconia/braze interfaces when the Cu 10 wt.% is used. For the case of the Cu 5T i wt.% alloy,
the expected reaction layer is T iO. According to the results presented above, the
only instance where the reaction product observed at the ceramic/metal interface
corresponds to what should be expected is when the Cu 5T i wt.% alloy is used at
9500 C. In the experiments corresponding to this case, it was observed that no metal
substrate dissolution into the braze.
In Chapter 6, it is shown that the nature of the metal substrate to be joined
to a ceramic may determine to a great extent the thermochemistry of the brazing
alloy. According to the analysis in that chapter, the large solubility of N i in Cu,
coupled to the highly negative interaction energy between N i and T i may result in
a dramatic decrease in the chemical reactivity of the brazing alloy. Because of the
high diusion rates within the liquid, the chemical activity of T i in the braze my be
131
1.0
BCC
BCC
0.9
BCC+HCP
0.8
Liquid+HCP
HCP
0.7
acr(Ti)
0.6
Liquid
0.5
Liq+M6X42
0.4
M6X33+M6X42
0.3
0.2
Liq+M6X33
M6X33+HCP
0.1
TiO
FCC
0.1
0.2
FCC+M6X33
0.3
0.4
0.5
0.6
0.7
0.8
FCC+TiO
0.9
1.0
x(O)/(x(o)+x(cu))
Figure 5-21: Activity diagram for the Cu T i O system at 10000 C.
reduced by a factor of 5 within 100 s from the start of the interdiusion reaction. In
the experiments presented above, the only case where the reaction product observed
at the ceramic/metal interface corresponded to what was expected was the brazing
experiment at 950C . According to the Cu T i phase diagram, at this temperature
and with the Cu 5 wt.% brazing alloy, the liquid fraction at the interlayer is less
than 50%. Under these conditions, the dissolution reaction is limited, and N i cannot
scavenge T i from the braze. At higher temperatures, however, N i dissolution occurs
and the chemical activity of T i in the braze is reduced below the stability range of
T iO. According to thermodynamic calculations in the Cu T i O system, in order
to observe T i2 O3 at the zirconia/braze interface, the chemical activity in the liquid
must decrease to values below 0.005. It is expected that when the ceramic/metal
joint is held at the brazing temperature for about an hour and the liquid fraction
in the interlayer is higher than 50% the chemical activity of T i should decrease to
0.005 or less. These observations have been conrmed in a related experimental work:
Sciti et al. [132] studied the brazing of Y SZ zirconia to Ni-based super alloys.
In their experiments, small pieces of zirconia and Ni-based super-alloy (thickness
2 mm) were assembled in a sandwich structure with a Ag 35.25Cu 1.75T i wt.%
brazing foil as an interlayer. The samples were brazed at 820 9500 C for 20 min.
and then cooled to room temperature. Their results are summarized below:
i) The reaction layer observed at the zirconia/braze interface was a T iOx oxide.
The exact stoichiometry was not determined, but it was established that the
composition of this phase was within the T i2 O3 < T iO2 range. This layer had a
thickness of about 1.2m.
132
ii) Immediately adjacent to this reaction product, a layer composed of N i, T i and

Cu was observed.
iii) It was also observed that both T i and Cu diused into the N i based super
alloy.
In the case of the brazing experiments with the Cu 10T i 5Al brazing alloy,
there was signicant T i scavenging. However, the observed reaction product at the
ceramic/metal interface was T i3 Cu3 O. To explain this, it is necessary to determine
the eect of Al on both the thermochemistry of the brazing alloy, and on the stability
of the T i3 (Cu, Al)3 O compound. In Section 4.3, a model was proposed to describe the
dissolution of aluminum in the T i3 Cu3 O M6 X compound. According to the calculations presented in that section, aluminum stabilized the M6 X structure and dissolved
in signicant amounts. Fig. 5-22(a) shows the eect of aluminum additions on the
stability of the T i3 Cu3 O compound. As can be seen, the minimum activity for which
this phase is stable decreases as the aluminum content in the braze increases. This
eect is coupled to the slight increase on the chemical activity of T i in Cu T i Al
alloys as the aluminum content increases (note that there is a maximum in the chemical activity of titanium as a function of aluminum content). Using these results, the
presence of T i3 Cu3 O at the zirconia/braze interface can be easily explained.
0.1
0.038
x(Cu)-40*x(Al)=0
x(Cu)-20*x(Al)=0
x(Cu)-13.333*x(Al)=0
x(Cu)-10*x(Al)=0
x(Cu)-8*x(Al)=0
0.09
0.037
0.07
0.036
acr(Ti)
Ti Activity, REf: BBC-Ti
0.08
0.035
0.06
0.05
0.04
0.034
0.03
0.033
0.02
0.032
0.01
0.02
0.03
0.04
x(Al)/x(Cu) Ra tio
0.05
0.06
0.07
0.01
0
(a) Minimum T i activity to observe

T i3 Cu3 O precipitation at ceramic/braze
interfaces as a function of aluminum content.
0.02
0.04
0.06
0.08
x(Al)
0.1
0.12
0.14
0.16
(b) Eect of aluminum on the activity of

T i in Cu T i brazing alloys for dierent
Cu T i ratios.
Figure 5-22: Aluminum eect on the thermochemistry of zirconia/CuT iAl interactions.
It is important to note that further evidence for the stabilizing eect of Al on

the T i3 Cu3 O phase comes from the experimental results on the active brazing of
alumina using Ag Cu T i-based brazing alloys. According to Kelkar et al. [72],
in all the experiments where alumina was joined to itself using Ag Cu T i brazes,
the sequence T iO/T i3 (Cu, al)3 O was observed at the alumina/braze interface. The
133
presence of aluminum, dissolved from the ceramic substrate, stabilized and promoted
the formation of the T i3 (Cu, al)3 O compound, in agreement with the Cu T i O
ternary activity map.
5.7
Interaction between Cu Zr melts and Zirconia
In all the experimental results described above, the reaction products observed at
the zirconia/liquid-metal interface formed a layered, approximately planar structure,
irrespective of the complexity of the reaction sequence. Once the reaction product
precipitated on the zirconia surface, the diusion of oxygen atoms across the reaction
products constituted a limiting factor. An entirely dierent interfacial structure could
be observed, however, if no solid phase precipitates on the zirconia surface and the dissolution of the zirconia substrate is thermodynamically favorable. Such an interface
is observed when CuZr brazes react with zirconia substrates, as shown in Fig. 5-23.
Figure 5-23: Cu 5 wt.% braze/zirconia interface, at 10500 C. Held at brazing temperature

for 1 hr.
The zirconia samples brazed in these experiments showed the typical blackening
observed in other experiments on zirconia/active metal interactions, indicating the
partial reduction of the zirconia lattice and the subsequent formation of oxygen vacancies and free charge carriers. From Fig. 5-23, it is evident that the Cu Zr braze
not only drives oxygen atoms from the zirconia lattice into the braze, but also attacks
the zirconia substrate. The irregular zirconia/braze interface clearly shows that some
zirconia particles partially detach from the bulk. In some cases, this detachment is
complete and zirconia particles can be observed to completely separated from the
zirconia surface. Unfortunately, there is almost no experimental information regarding the thermodynamics of the Cu Zr O (see Section 4.4), and the approaches
outlined above cannot be used. Despite this, it is possible to calculate whether, under
134
these thermal and chemical conditions, extensive zirconia dissolution in the melt can
be expected. To do this, the approach outlined in Section 2.3 can be followed.
The dissolution of ZrO2 into a Cu Zr melt (ZrO2 (Zr) + 2(O)) can be
studied using Eq. 2.6 in 2.3.1:
aZr a2 = exp
O
GZr
f
RT
xZr Zr (xO O )2 = exp
(5.2)
GZr
f
RT
where
ln (Zr ) = ln (Zr ) + Zr xZr + Zr xO

o
Zr
ln (O ) = ln (O ) + O xO + Zr xZr
o
O
Assuming that Zr xZ r and Zr << lnZ r and O xO 0 (see Table 5.9):

Zr
O
O
GZrO2
f
RT
xZr (xO )2 exp 2Zr xZr = exp

O
Zr (O )2
If there is no Zr initially in the melt, it is possible to use the relationship xZr =

and we have (since xZr 0):
xa
O
GZrO2
f
RT
2 exp
Zr
(O )2
xO
2
1/
3
(5.3)
When there is already Zr in the melt, xZr >> zO , and therefore

xb
O
exp
GZrO2
f
RT
Zr
1
1
Zr
2 x
(O )
Zr exp (2O xZr )
1/
2
(5.4)
Using the thermochemical values for Cu Zr O melts at 11000 C in Table 5.9,

it is possible to calculate the two dierent cases presented in Eq. 5.3 and Eq. 5.4,
considering an initial zirconium concentration for the second case of xCu = 0.05. The
calculations yield xa = 4 1010 and xb = 9 1012 . As expected, xa >> xb .
O
O
O
O
Based on the analysis presented above, one would expect the dissolution of ZrO2
into Cu alloys to be signicantly decreased when Zr is added to the melts. Obviously,
this is not what has been observed in the experimental results, exemplied by Fig. 523. This discrepancy between theory and experiments shows that there is a limit to
the simplied analyses presented in this work. In this particular case, the observed
interfacial structure cannot be explained using the bulk thermodynamic properties of
the phases involved. Although explaining the nature of this interfacial reaction lies
135
Table 5.9: Thermochemistry of Cu Zr O melts at 11000 C in the Cu-rich region.
GZrO2
f
836, 971 J/mol(ZrO2 )
ln (O )
2.03
ln (Cu )
3.7
O
O
8.57
Zr
Zr
0.6
Zr
O
118
O
Zr
118
beyond the scope of this work, it may be possible to present a hypothesis regarding
the apparent attack of the zirconia substrate by Cu Zr melts.
In general, defects (one or two-dimensional) constitute highly energetic and chemically active regions within a phase. It is likely then that defects within the zirconia
substrate, such as grain boundaries, were selectively attacked by the Cu Zr melt.
Initial depletion of oxygen within these grain boundaries may have made them unstable with respect to further decomposition and dissolution into the attacking liquid.
This attack may have been aided by liquid permeation through the porous zirconia
structure. Once the grain boundaries were selectively attacked, grains or particles of
the zirconia substrate could be easily detached from the bulk.
This hypothesis, however, is highly speculative, and further analysis of the interfacial structure is required. Furthermore, in order to rationalize the formation of this
interface requires a much more sophisticated theory, which is denitely beyond the
scope and objectives of this thesis.
5.8
Application of Thermodynamics to the Study of Zirconia/Metal Interfaces: Conclusion
In this chapter, the thermodynamic models developed in Chapter 3 and Chapter 4

have been used to rationalize several interesting interfacial phenomena occurring
when zirconia-based ceramics are put in contact with reactive metals. In the case
of ZrO2 /T i interactions, it was shown how the use of metastable thermodynamic
calculations allow the prediction of reaction paths that are in good agreement, at
least qualitatively, with reported experiments. According to the analysis presented in
this chapter, the relative size of the zirconia and titanium subsystems inuences the
reaction path observed. The interactions between active brazing alloys and zirconia
substrates were analyzed using chemical activity diagrams and simple thermodynamic
calculations. From these analyses, it was concluded that in the majority of the cases
136
the reaction layers formed at the interface satised the local equilibrium condition,
despite the relatively short brazing times in some ceramic/metal joining experiments.
137
138
Chapter 6
Metal Substrate Eects on the
Thermochemistry of Active
Brazing Interfaces
6.1
Introduction
In most cases, ceramics are not wetted by molten metals [9]. The reason for this is the
relatively high stability of ceramic surfaces: a ceramic/vapor interface is more stable
than a ceramic/liquid interface. Therefore, in ceramic/metal (C/M ) joining applications using a liquid interlayer between the ceramic and metal parts (i.e. brazing), the
melt usually has to be chemically modied to promote wetting. Adding chemically
active elements to the melt improves the wettability of molten metals over ceramic
substrates for two reasons (see Section 2.6):
i) The active element can promote a greater ceramic dilution. Dissolved oxygen
reduces the Solid/Liquid interfacial energy and increases the work of adhesion of
the C/M system.
ii) If the interaction between the active element and the anion is strong, the precipitation of a reaction layer can be promoted. If this reaction layer has the
appropriate electronic properties wetting is further promoted and the work of
adhesion increases to levels that make C/M interfaces practical.
Much of the literature on C/M active brazing has been focused on the nature of
the interactions between the ceramic substrate and active brazing alloys (see for example Eustathopoulos and Drevet [16]). Almost nothing, however, has been published
on the analysis of the eects of using dierent metal substrates on the thermochemistry of the brazing alloys themselves. It will be shown in this work that the metal
substrate can have a signicant eect on the thermochemical behavior of the brazing
alloy.
i.e. metallic in character.
139
The chemical activity of the active element used in the brazing alloy determines
the extent and character of interfacial reaction at the C/M interface . Therefore,
studying the changes in the reactivity of the brazing alloy as it interacts with the
metal substrate is of great importance when designing and analyzing C/M active
brazing processes. In this chapter, the work by Arroyave and Eagar [133] on the
metal substrate eects on the thermochemistry of brazing alloys will be presented
and complemented with some experimental results by Arroyave [51].
6.2
Previous Experimental Evidence
There are a number of experimental results available in the literature that illustrate
the eects of using a particular substrate material on the thermochemistry of active
brazing alloys. Through their work on alumina/metal joining processes, Barry and
Leatherman [134] found that the brazing temperature had a noticeable eect on the
nal tensile strength of alumina/(Fe-29Ni-17Cu) joints. Using three dierent Cubased active brazing alloys they found that by brazing at temperatures much higher
than the alloy liquidus, the strength of the joint was greatly decreased. As the temperature increased, a larger N i content was observed and this correlated with a lower
tensile strength.
Stephens et al. [135] found that when brazing alumina to Kovar r with Ag
CuT i alloys the T i active element was scavenged from the brazing alloy and N iT i
intermetallic compounds were formed (N i3 T i was the most usual intermetallic phase
observed). The scavenging prevented the formation of a continuous T ix O layer at
the alumina/braze interface. It was found that the joints with discontinuous reaction
layers oered poor hermeticity. By using a M o-based diusion barrier, it was found
that dissolution of the Kovar r substrate was decreased and a continuous T ix O layer
was observed at the alumina/braze interface. The performance of the joints using
this diusion barrier was noticeably better that those without it.
The detrimental eect of N i dissolution into the braze alloy is not limited to
ceramic oxide/metal joining applications. Kim et al. [136] investigated silicon nitride/steel joints using N i-based interlayer materials and T i-based active brazing
alloys. In their work, Kim et al. [136] found that the N i interlayer dissolved into the
brazing alloy. As this dissolution increased, the joint strength decreased due to the
formation of a brittle N i T i compound and a decreased reactivity of the braze at
the interface with the ceramic. Due to the decreased reactivity of the brazing alloy,
the active brazing alloy did not wet the ceramic substrate.
Using the same type of active brazing alloys used in the experiments previously
described, Bucklow [137] studied zirconia/cast iron joints. In this case, no detrimental excessive dissolution reactions between the substrate metal and the brazing alloy
were reported.
Determining, for example, the interfacial reaction product.
140
Based on the experimental evidence above, complemented by the experiments described in Section 5.6 it is possible to conclude that:
At temperatures much higher than the liquidus of the CuT i brazing alloys, the
reaction between the brazing alloy and the ceramic is negligible or non-existent
when the metal substrate has a high N i content.
The same eect is less pronounced when the alloys main component is F e.
In brazing experiments with N i-based metal substrates, considerable N i dissolution in the brazing alloy is observed. Moreover, when N i is dissolved in great
amounts, brittle N i T i precipitates have also been observed.
6.3
Substrate-Brazing Alloy Interaction: Simplied Analysis

Table 6.1: Enthalpies of innite dilution of T i in melts.
Solvent
Enthalpy of Dilution, J/mol
Cu
10, 000 [30]
Fe
54, 000 [138]
Ni
187, 000 [139]
Table 6.1 shows the calculated partial enthalpies at innite dilution for T i in different liquid solvents. As can be seen, the innite dilution enthalpy for T i in N i is
very negative and therefore it should be expected that the chemical activity of T i in
N i melts has very large negative deviations from ideality .
At 1300 0 C, and using the thermodynamic description for the N i T i binary system according to Kaufman [139]:
0
RT ln T i = 102, 000 J/mol
(6.1)
N
Therefore, T ii 4 104 . Thus, at innite dilution the activity of T i is greatly
reduced if N i is present in signicant amounts. In the case when T i is dissolved into
F e, under the same dilution and temperature conditions and using the thermodyF
namic description given by Jonsson [138], T ie 102 . Thus, the eect of F e on the
chemical activity of T i should be, based on these simple calculations, much smaller
than that of N i.
i.e. large negative interaction energies

<< 1
141
1500
1800
1450
1600
1400
1400
Temperature, 0C
Temperature, 0C
Liq
bcc
Liq
1350
1300
1250
fcc
1200
fcc
1200
1000
fcc
1150
800
1100
1050
0.2
0.4
0.6
Mole Fraction x(Ni)
0.8
1.0
(a) Cu N i phase diagram.
600
bcc
0
0.2
0.4
0.6
0.8
Mole Fraction, x(Cu)
1.0
(b) Cu F e phase diagram.
Figure 6-1: Calculated Cu N i and Cu F e phase diagrams.
Using thermodynamic arguments, it is also be possible to examine the nature of

the interactions between Cu and the metal substrates N i and F e.
The lenticular shape of the Cu N i phase diagram in Fig. 6-1(a) shows that
the system is nearly ideal. The system exhibits complete solubility in both the solid
(f cc) and the liquid phases. It is therefore reasonable to expect that, when in contact
with a Cu-rich liquid, signicant amounts of N i will be dissolved in the melt. The
Cu F e phase diagram (Fig. 6-1(b)), on the other hand, exhibits very little solubility
of F e into Cu-based liquid solutions, so the degree of dissolution of an F e-rich metal
substrate into a Cu-based brazing alloy should be signicantly less.
From this analysis, it follows that N i will have a much larger eect than F e on
the chemical activity of T i in Cu melts due to two factors:
There is a very strong anity between N i and T i (see Table 6.1).
N i dissolves extensively in Cu melts (see Fig. 6-1(a)).
6.4
Thermochemical Analysis of Metastable Interfacial Systems
To verify the analysis presented above, numerical simulations of the thermochemical

interactions between the two dierent metal substrates and Cu-based active brazing
alloys were performed. For simplicity, the analysis was focused on two commonly
used metal substrates (N i and F e) and a simple active brazing alloy belonging to the
Cu T i system (95 at. %Cu). This analysis, however, can be applied to other active
brazing/metal substrate combinations that are important in ceramic/metal joining
applications. Thanks to recent developments in computational thermodynamics, it is
now possible to perform complex stable and metastable thermodynamic calculations
142
involving large numbers of species and phases [21]. These calculations involve the
use of suitable thermodynamic models for each of the phases participating in the
calculated equilibria. In this work, CALPHAD thermodynamic descriptions for the
Cu N i T i and Cu F e T i systems were used to analyze the nature of the metal
substrate/braze interactions.
By integrating previously assessed thermodynamic data [30, 138141] with readily available experimental diusion coecient measurements [121] it was possible to
simulate the diusional reactions taking place at the metal-substrate/brazing alloy
interface just after it is assembled and subjected to elevated temperatures. These
simulations involved the calculation of a (one-dimensional) moving boundary-type
problem with a metastable solid/liquid interface so it was necessary to use metastable
thermodynamic extrapolations for the systems of interest. The thermodynamic calculations in this work were performed using Thermo-Calc r [142].
Within the explored temperature range (1000 0 C 1300 0 C) both N i and F e in the
pure state exist in the f cc crystal structure. Therefore, all the calculations necessary
to compare the eect of these two elements on the chemistry of active brazing alloys
involved only the f cc and liquid phases. In the present analysis, it was additionally
assumed that the kinetics for the precipitation of binary and/or ternary intermetallic
compounds is very sluggish compared to the diusion rates of the elements at the
S/L interface. Eectively, this excluded the precipitation of such phases from the
analysis, which is a reasonable assumption for calculations involving the very rst
stages of the dissolution of the metal substrate into the brazing alloy .
6.5
Kinetic Analysis of liquid/f cc Moving Boundary.
The diusion in a multicomponent system can be expressed by the Fick-Onsager law:

n1
n
Dkj Cj
Jk =
(6.2)
j=1
where Jk is the diusional ux of species k and Cj is the concentration (in moles/m3

n
), while Dkj is the diusion coecient in the (n 1) by (n 1) diusion matrix [143].
When there is a diusional reaction between two phases, say and , the velocity
v of the moving boundary can be calculated using the ux balance equation:
v =
Jk Jk
Ck Ck
for k = 1 . . . n 1
(6.3)
where Jk and Jk are the mass uxes in the and phases at the interface , and
After long periods of time, it can be expected that those phases with positive driving forces for
precipitation would start nucleating.
143
Ck and Ck are the corresponding interfacial concentrations.
A two-phase equilibrium in an n-component system has n 2 degrees of freedom at a xed temperature and pressure. Therefore the same number of constraint
equations should be found in order to determine the tie-line, i.e. the iso-activity line
linking the two equilibrium compositional coordinates at each side of a two-phase
eld. The tie-line denes the interface compositions used in Eq. 6.3. This equation
can be applied to any of the solute components. In a ternary system there are two
solute components and therefore there are 2 equations of the type of Eq. 6.3. The
interfacial velocity must be the same no matter which solute component is chosen
and therefore:
J1 J1
C1 C1
J2 J2
C2 C2
(6.4)
In a ternary system, it is necessary to solve one non-linear dierential equation

(Eq. 6.4) in order to nd the correct tie-line, i.e., the interfacial composition of the
two phases. Once the interfacial compositions are known by nding the appropriate
tie-line through the mass conservation equation (Eq. 6.4), Ficks second law can be
solved for each of the regions in the simulation:
Ci
= Ji
(6.5)
t
The interfacial compositions obtained from solving Eq. 6.4 become the boundary
conditions necessary for solving Eq. 6.5.
The simulations were performed in the commercial software DICTRA r [144],
which is capable of performing local equilibrium calculations and diusion simulations for moving boundary problems. The kinetic data were taken from Smithells
[121]. Whenever there was no data for the impurity diusion coecient of one of
the elements into another, the self-diusion coecient of the element acting as the
solvent was used as a rst approximation. For the case in which T i had to be treated
as the solvent element in the (metastable) f cc phase , the diusion coecients were
taken from the data measured for the hcp phase, given the fact that both phases are
closed packed and therefore would present a similar kinetic behavior . The values
for the diusion coecients of all the elements in the liquid phase were set to be
2
1 109 m , according to criteria set by Kirkarldy and Young [122] .
s
6.6
Simulation
To analyze the thermochemical interactions occurring at metal-substrate/brazing alloy interfaces the simulation was constrained to a simple one-dimensional geometry.
Since the mobility for the hcp is non-isotropic an average value considering the mobility values
parallel and perpendicular to the basal plane was used.
144
The simulation domain was composed of two distinct regions. The region on the
left-hand side was dened as a liquid phase (Cu T i brazing alloys) and a width of
100 . The right-hand side was dened as an f cc phase (either pure F e or N i) and
a width of 1, 000 :
100 microns
Liquid Braze Alloy
1000 microns
Solid Base Metal
Figure 6-2: Schematics of the simulation domain.
The number of grid points in each computational region were allowed to be varied
as the simulation progressed. A higher density of grid points was used on both sides
of the S/L interface.
Zero ux conditions were used on the left and right boundaries of the computational domain and the simulated time for all the cases explored was 100 s, although
simulations for much longer times (t > 106 s) were performed to verify that the system
reached the expected equilibrium state. The initial conditions for the liquid region
are the same for both the Cu N i T i and Cu F e T i subsystems (5 at. %T i,Cubalance.) The initial conditions for the solid regions were either pure F e or pure N i
in the f cc structure.
To determine the nal state of the closed system for the given initial conditions,
metastable equilibrium calculations were run and it was found that the metastable
nal state is f cc in all of the cases. To verify that the systems studied reached the
equilibrium state, the diusion simulations were run for times in excess of 106 s for
both the Cu F e T i and Cu N i T i cases. In Fig. 6-3 it is possible to see
the simulation results for long times for both the N i/(Cu T i) and F e/(Cu T i)
systems . Note that these calculations are not really applicable for long times after
the diusion reaction begins, as some of the phases whose precipitation is limited by
kinetic factors would then have plenty of time to nucleate and grow.
From these simulations, it is possible to extract the following results:
The system was closed.

The simulations for the Cu F e T i system at 11000 C and 12000 C were so sluggish that the
equilibrium state was not reached until t > 1010 s.
145
25
13000C
Position of Sol./Liq. Interface, m x 10-5
Position of Sol/Liq. Interface, m x 10-5
25
20
20
15
10
15
12000C
11000C
13000C
10
0
10-3 10-2 10-1 100 101 102 103 104 105 106
Time, s
(a) S/L interface position calculations

for the Cu N i T i system.
11000C
0
10-4
12000C
10-2
100
102 104
Time, s
106
108
1010
(b) S/L interface position calculations

for the Cu F e T i system.
Figure 6-3: S/L interface position calculations.
At the rst stages of the dissolution reaction, the solid region begins to dissolve
into the liquid. The extent of this dissolution is greatest in the case of the
N i/(Cu T i) interaction, due to the much greater solubility of N i into Cu
melts.
After the initial enlargement of the liquid region (larger when the metal substrate is N i-based), diusion of both Cu and T i into the solid f cc phase takes
place and the liquid layer starts to shrink
If these metastable inter-diusion reactions were to continue indenitely, the
layer would solidify isothermally, provided that the other intermetallic compounds did not overcome the kinetic barriers for nucleation and subsequent
growth.
From Fig. 6-3 it is evident that it takes at least two orders of magnitude longer
for the liquid in the F e/(Cu T i) diusion reaction to disappear.
For both system, the enlargement of the liquid layer reaches a maximum at
around 100 s.
In order to make a proper comparison between the eects of N i and F e on the
thermochemistry of brazing alloys it was necessary to nd a proper time at which
both systems were in equivalent states. As seen in Fig. 6-3 the N i and F e cases reach
equilibrium, meaning disappearance of the liquid layer, at very dierent times . It
is worth noting that once T i and Cu start to diuse into the substrate metal, the
A dierence greater than two orders of magnitude
146
chemical activity of T i not only decreases because of the chemical interaction with
either N i and F e in the liquid layer, but because the actual amount of T i in the liquid layer decreases as well. Additionally after very long times, it is likely that phases
with positive driving forces for precipitation would start to nucleate and grow. These
phases would necessarily be dierent in the two cases. Thus, the equilibrium states of
the systems do not constitute a proper common ground for making the comparison.
From Fig. 6-3 it is seen that maximum dissolution is reached in both systems at
similar times ( 100 s). The point of maximum dissolution was therefore selected as
the common state at which a comparison between both systems was meaningful.
0.08
0.18
i)
tio
n
rac
5s
0.10
100 s
0.08
Mo
0.04
0.01
0
0.12
0.06
5s
0.02
0.16
0.14
le F
Solid
100s
0.03
x(T
tio
nx
(Ti
rac
le F
Mo
0.06
0.05
0.04
0.20
Solid
0.07
50s
0.2
0.4
0.6
0.8
Mole Fraction x(Cu)
5s
0.02
Liquid
1.0
(a) Diusion paths for the Cu N i T i

system.
0.2
0.4
0.6
0.8
1.0
Mole Fraction x(Cu) Liquid
(b) Diusion paths for the Cu F e T i

system.
Figure 6-4: Diusion path calculations.
Fig. 6-4 shows the calculated diusion paths for both the Cu N i T i and
Cu F e T i systems (at 13000 C) for very short times after the inter-diusion reactions begin. The diusion paths are superimposed over the calculated metastable
ternary phase diagrams at the same temperature. At very low T i contents, it is expected that both ternary systems exhibit behaviors similar to those of the dominant
binary systems, Cu N i (see Fig. 6-1(a)) and Cu F e (see Fig. 6-1(b)). Fig. 6-4(a)
and Fig. 6-4(b) show that the Cu N i T i ternary system exhibits a much wider
range of solubility for both the liquid and f cc phases than the Cu F e T i ternary.
The diusion paths shown in Fig. 6-4 show that after very short times (t < 50 s)
the composition of the entire liquid region approaches the composition given by the
tie-line corresponding to the metastable f cc/liquid equilibrium condition at any given
time. This liquid composition is indicated by the lled square in both diagrams. Because the diusion rates at the S/L are slow compared to the diusion rates in the
liquid phase, it can be expected that the liquid composition will always be homoge
The point at which the enlargement of the liquid layer reaches a maximum
147
neous after the rst 100 s of the reaction and the composition of the liquid will
thus follow the liquid/(fcc+liquid) phase boundary.
In both systems, the T i and Cu content of the liquid phase will start to decrease
as the diusion reaction progresses (this is seen more clearly in Fig. 6-4(b)). The
diusion paths will thus follow decreasing T i activity tie-lines (in the direction of the
arrow) until the liquid phase disappears and the nal composition in the sample is the
one given by the initial conditions of the simulation (xCu 0.08 and xT i 0.005). It
can be seen in Fig. 6-4(a) that for both systems, the nal composition (at 13000 C)
indicates that the liquid should disappear entirely (the nal state of the system is
indicated by a triangle), as can also be seen in Fig. 6-3.
Fig. 6-5 show the metal substrate mole fraction (at.%) proles for both simulated
systems at dierent temperature conditions and at dierent times. It is evident that
in all cases, after relatively short times ( 50 s) the molar concentration proles for
the metal substrate become at in both regions (solid and liquid) of the simulation
domain.
In the liquid region, the diusional processes are so fast that once the system
reaches a tie-line the composition of the liquid is more or less uniform and equal to
the composition at the metastable solid/liquid interface throughout the region (see
also Fig. 6-4). Although the liquid composition is uniform throughout the entire liquid layer, this does not mean that the composition remains constant during the entire
reaction. Slow, not fast, diusion rates are the main cause for the relative atness of
the substrate metal (N i or F e) composition prole in the solid region: the diusion
of both Cu and T i into the metal substrates is so slow that at short times there is
virtually no diusion whatsoever. Thus, just after the diusional reaction begins,the
mass ux direction across the S/L interface is from the substrate metal towards the
melt. After long enough times, diusion across the solid region becomes important
and the liquid starts to disappear as both Cu and T i diuse out of the liquid and
into the substrate metal.
From Fig. 6-5(a) it is seen that at 13000 C the dissolution of the N i metal substrate is extensive and the molar concentration of N i in the Cu T i brazing alloy at
the point of maximum dissolution is relatively high ( 50 at.%). At the same temperature conditions, the dissolution of F e into the Cu T i brazing alloy is 5 times
lower (see Fig. 6-5(b)).
From these results and the values for the N i T i and F e T i chemical interaction energies given in Table 6.1, it may be expected that the nal chemical activity
of T i in the Cu liquid brazing alloy after 100 s will depend on the nature of the metal
substrate.
That is, a two-phase local equilibrium condition at the solid/liquid interface.
148
TIME = 0,5,50,100 s
0.6
Liq
Sol
Liq
Liq
Sol
Sol
Mole Fraction x(Ni)
0.5
50,100 s
0.4
0.3
50,100 s
0.2
50,100 s
0.1
5s
0
0s
0s
10
5s
5s
15
20
-5
25 0
10
0s
15
20
Distance, m x 10
Distance, m x 10
25 0
-5
10
1200 C
1100 C
15
20
25
-5
Distance, m x 10
1300 C
(a) xN i prole for the Cu N i T i system.
TIME = 0,5,50,100 s
0.6
Liq
Sol
Mole Fraction x(Fe)
0.5
Liq
Liq
Sol
Sol
0.4
50,100 s
0.3
0.2
50,100 s
50,100 s
0.1
0
0s
0
15
20
5 s 10
-5
Distance, m x 10
1100 C
0s
5s
25 0
10
0s
5s
15
20
Distance, m x 10
-5
25 0
1200 C
10
15
20
-5
25
Distance, m x 10
1300 C
(b) xF e prole for the Cu F e T i system.
Figure 6-5: Molar fraction prole for the diusion couple calculations.
Fig. 6-6 shows the chemical activity of T i 100 s after the substrate/braze reaction
begins for three dierent temperature conditions and for F e and N i-based substrates.
Because of the greater solubility of N i in Cu and the strongly negative chemical interaction between T i and N i the eect of N i on the thermochemistry of the braze is
greater than that of F e. When F e dissolves into the braze (after 100 s), the activity
of T i decreases to 70% of the level it had at t = 0 s and T = 11000 C. At the same
temperature and with the same elapsed time, N i causes the T i chemical activity to
decrease to 20% of the initial value. The eect becomes even more noticeable when
comparing both cases at 13000 C: while the T i activity levels in the Cu F e T i
system have decreased to 40% of the original value, the T i activity in the CuN iT i
system is so low (< 1%) that the brazing alloy ceases to be active in the strict sense
of the word.
The chemical activity is normalized with respect to that found in the braze at t = 0 and
T = 13000 C.
149
acr(Ti) Normalized with respect to t=0s, T=1100 C
1.4
t=0
100s,Ni
100s,Fe
1.2
0.8
0.6
0.4
0.2
0
1100
1120
1140
1160
1180
1200
Temp, 0 C
1220
1240
1260
1280
1300
Figure 6-6: Normalized chemical activities of T i at t = 0 and t = 100 s for the N i and
F e cases. The activities are normalized with respect to the chemical activity of T i in the
Cu T i braze at t = 0 and T = 1000 C
As noted in Section 6.2, several authors have found that one of the results of
an excessive reaction between the braze alloy and a N i-rich metal substrate is the
formation of brittle N i T i compounds [135, 136]. Using the calculated liquid composition obtained after 100 s for the Cu-Ni-Ti case at 12000 C (see Fig. 6-5(a)) the
equilibrium phase fraction as a function of temperature was calculated. According to
this calculation, shown in Fig. 6-7, the N i T i intermetallic compound that is likely
to form (scavenging signicant amounts of T i from the brazing alloy) is N i3 T i, as
reported by Stephens et al. [135].
6.7
Experimental Verication: Zirconia/Cu-Ti/Ni Interactions
As has been noted in Section 6.2, sometimes the reactions observed in ceramic/metal
joining operations do not correspond to what could be expected when the analysis
only incorporates the thermodynamics of the brazing alloy and the ceramic. The
discrepancies between experiment and prediction can often be attributed to the eect
of the metallic substrate on the thermochemistry of the brazing alloy used in the
C/M joining operation. As has been seen in this section, metal substrates rich in N i
can have specially strong eects on the thermochemistry of Cu-based brazing alloys.
Because of the strong N i T i chemical interaction, if N i is substantially dissolved
in the Cu-based alloy, the T i chemical activity will be decreased. This eect is dramatically increased as the temperature at which the brazing operation is performed
increases.
Arroyave [51] studied the interfacial reactions occurring at zirconia/braze interfaces when the ceramic was joined to a N i-based (Inconel 718 r ) super alloy, using
150
(x(Ni)~0.3 x(Ti)~0.03)
1.0
0.9
Solid
0.8
Liquid
Phase Fraction
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Ni3Ti
0
500
1000
1500
Temperature, 0C
2000
Figure 6-7: Phase fraction calculation for CuN iT i case, after t = 100 s and T = 12000 C.
(a) ZrO2 /Cu T i/Inconel 718 r joint

created at 10500 C and held at that temperature for 1 hr
(b) ZrO2 /Cu T i/Inconel 718 r failed

joint created at 11500 C and held at that
temperature for 1 hr
Figure 6-8: ZrO2 /Cu T i/Inconel 718 r joints brazed at two dierent temperatures [51].
Zirconia substrate is at the bottom of both micrographs.
Cu-based active brazing alloys. According to the model developed in this section, a
noticeable eect should be observed in the nature of the interfacial reactions occurring
at the zirconia/braze interface. In fact, if the temperature is increased many degrees
above the liquidus line, N i dissolution from the metal substrate into the liquid active
brazing alloy should be so extensive, that the brazing alloy could be rendered chemi151
cally inactive, at least for practical purposes.

Fig. 6-8 presents a dramatic verication of the model presented in this section:
Fig. 6-8(a) presents a micrograph of a zirconia/Cu 10 wt.%T i/Inconel r joint
created at 1, 0500 C. As can be seen in the gure, a reaction layer of T iOx composition
is present at the zirconia/braze interface. The titanium activity was thus high enough
to drive oxygen ions from the zirconia lattice through the zirconia/braze interface.
Oxygen then reacted with the titanium dissolved in the braze. The reaction layer
thus formed was then wetted by the liquid braze and a continuous ceramic/metal
interface was created.
When the same zirconia/Cu 10 wt.%T i/Inconel r arrangement is brought up
to increasingly high temperatures, nickel migration from the Inconel r substrate
increases, as has been shown in Section 6.6. Fig. 6-8(b) shows an extreme case
of dissolution. This sample was heated to 1, 1500 C and, as can be seen from the
micrograph, the braze almost completely dissolved the Inconel r substrate. The titanium activity was therefore reduced to levels at which no reaction layer was formed
at the ceramic/braze interface and no continuous ceramic/metal interface was therefore formed. According to the model presented in this section, maximum dissolution
should take place in the order of 1, 000 s. The sample was held at 1, 1500 C for 3, 600 s
and the model has been validated.
Even at the relatively low temperature at which the joint shown in Fig. 6-8(a) was
created, there should be some nickel dissolution. Due to the high chemical interaction
between nickel and titanium, even such reduced amounts of nickel should be enough
to eectively scavenge some of the titanium that otherwise would be available to react
with the oxygen driven from the zirconia lattice. Fig. 6-9 shows an enhanced x-ray
map of the sample shown in Fig. 6-8(a). As can be seen, the regions rich in nickel
and titanium almost overlap completely. This provides additional verication of the
scavenging eects explained in this section.
6.8
Metal Substrate Eects in Ceramic-Metal Joining: Concluding Remarks
In the present section the nature of the thermochemical interactions between a metal
substrate (F e and N i) and an active brazing alloy (Cu T i) at short times at high
temperatures have been investigated. Based only on the enthalpies for innite dilution of T i in F e and N i it was concluded that N i would have a much stronger eect
than F e when they were in solution in a T i-containing Cu brazing alloy. From phase
diagram calculations it was evident that N i is much more soluble than F e in Cu melts.
To gain a deeper understanding of these eects, we proceeded to analyze the na
The sample was digitally processed and the color codes were enhanced to make the titanium
scavenging by nickel more evident.
152
Figure 6-9: X-Ray mapping of ZrO2 /Cu T i sample. Sample held at 10500 C for 1 hr [51].
ture of the thermochemical interactions occurring in the brazing alloy/metal system.

In order to do so, metastable extrapolations for the assessed systems were used to
simulate a moving-boundary problem in an f cc/liquid system. The dissolution of
the metal substrates (F e and N i) in the brazing alloy was calculated and resulting
eect on the chemical activity of T i in the melt was determined for several times
and temperatures. Through this analysis it was found that both metal substrates
decrease the chemical activity of T i in the melt. However, the eect is much more
intense when the metal substrate is N i-based.
The nature of the reaction products and extent of the ceramic/metal reaction can
be modied by changing the T i chemical activity of the brazing alloy. It has been
proven in this section that the metal substrate in a C/M joint can have a great eect
of the reactivity of the brazing alloy. Thus, the metal used in a C/M joint and the
actual joining temperatures are factors that need to be considered when designing a
C/M active brazing procedure.
153
154
Chapter 7
Using Phase Field Techniques to
Model Coupled Oxide Growth
during Active Brazing of Ceramic
Oxides
7.1
Introduction
As noted in Section 2.6, several industrial applications have required the use of novel
ceramic/metal C/M composite parts. In order to obtain reliable C/M joints, several
techniques for joining these materials have been developed [see 1]. One of the most
commonly used techniques for joining ceramic oxides (such as ZrO2 , Al2 O3 ) and metals is reactive brazing, which relies on the use of conventional brazing materials (such
as Cu Ag alloys) with active additions (such as T i) to promote an exchange reaction in which the reactive element is oxidized by the oxygen present in the oxide (see
Section 2.6). As time proceeds, the interfacial products formed at the C/M interface
evolve and grow, aecting the reliability and long-term behavior of the joints created.
It is therefore important to develop reliable models that describe the formation and
growth of reaction products during C/M bonding applications.
The modelling of C/M joining processes requires a thorough knowledge of the
thermodynamic and kinetic behavior of the elements and phases involved. Several
models have been proposed for dierent C/M systems. In a few cases, the models
describe describe the growth of only a single transformation product . Thus, relatively simple one-phase, one-mechanism diusion models can be used. Unfortunately,
for the majority of the cases relevant to practical applications, the reaction at C/M
interfaces involves the formation and subsequent growth of a series of reaction products [146], whose layering and morphology is determined by both thermodynamics
and kinetics. In most cases, the usual approach to analyze these interfacial reactions
consists of splitting the various possible diusional transformations into individual
For example, in the study of the reaction between liquid silicon and carbon studied by Zhou
and Singh [145], the only reaction product considered is SiC.
155
stages, which are then separately analyzed and then coupled in such a way that equilibrium and conservation conditions (both local and global) are satised.
In this chapter, a model based on the Cahn-Hilliard equation [147, 148] for the
description of the kinetic evolution of a heterogeneous system with diuse interfaces
is described. The particular problem that will be addressed is the oxide layering and
growth occurring when zirconia-based ceramics are reactively brazed to a metal using
T i-containing brazing alloys. Thermodynamic and kinetic factors will be discussed
and the details of the numerical model will be commented.
7.2
Interactions between Zirconias Ceramics and Ti-containing Brazes
As has been noted above, zirconias (ZrO2 ) constitute a very important family of
ceramic oxides used in several industrial applications because of their great thermal
stability, low thermal conduction, ionic transport properties, and so forth. In many of
this cases, the joining of these ceramic to metals is essential for the proper functioning
of the particular application. In order to create these ceramic/metal, C/M, composite parts using liquid-based joining approaches , it is necessary to incorporate active
element additives (such as T i) to promote wetting on the otherwise non-wettable zirconia surfaces . As noted in Eustathopoulos et al. [5, chap. 6], wetting of ceramic
substrates by molten metals is promoted by active additions due to either the formation of a wettable reaction product, the reduction of the C/M interfacial energy
due to surface co-segregation of the active element (T i, for example) and the anion
(O in this case). As zirconia reacts with T i , a pattern of temper colors (blackening)
develops in the zirconia ceramic. This pattern begins to be observed at the zirconia/titanium interface and, as the reaction progresses, the blackened region expands,
sometimes covering the entire thickness of the ceramic substrate [51]. The blackening, as has been observed by Park and Blumenthal [101], Janek and Korte [105] is a
qualitative indication of an increased interaction between photons and free carriers,
whose formation can be explained by the reaction:
where VO
vacancy).
1
X
X
2ZrZr + OO 2 ZrZr + VO + O2 (g)

(7.1)
2
represents a doubly-ionized oxygen vacancy (or free electrons within the
For example brazing, transient liquid phase diusion bonding, etc.

As has been noted by Li [9], a high thermodynamic stability of the ceramic substrate is related
to a very low work of adhesion for the interface between a non-reactive liquid metal and the ceramic
itself.
The reason for this segregation is that the perturbations in the structure and energy of bulk
metals due to the presence of electronegative elements such as oxygen can be better relaxed in
2D geometries [5]. Due to the high chemical anity between the active element and oxygen, it is
reasonable to expect that the active element will co-segregate to the free surface.
Either present as a thin lm or dissolved in the liquid braze.
156
Because of the high chemical anity between O and T i , a large oxygen chemical
potential gradient is created across the C/M interface as soon as the ceramic is put in
contact with the titanium layer. This gradient constitutes the driving force necessary
to carry out Eq. 7.1. As oxygen vacancies are created, oxygen atoms diuse towards
the zirconia/T i interface. As zirconia reacts with T i, a sequence of titanium oxides
(see Fig. 7-1) is formed at the interface and the complex layering grows with time
until thermodynamic equilibrium is reached.
Ti4O7
TiO
3
hcp
bcc
ZrO2
(O) Ref:O 2,g
TiO2
Zr O
Ti O
Ti3O5
x 10
Ti2O
hcp
5
bcc
6
7
8
9
0.1
0.2
0.3
x(O)
0.4
0.5
0.6
0.7
Figure 7-1: (O) vs. xO for the ZrO and T iO systems. Calculated using thermodynamic
models by Arroyave et al. [52] and Lee [34].
According to experimental results [51, 149] the titanium reaction layer often found
at zirconia/T i interfaces is T iO1x . The nature of the rst-forming titanium-oxygen
compound at the C/M interface is determined by the interfacial oxygen chemical potential , which can be xed by additional chemical reactions occurring in the system.
For example, the titanium oxide present at zirconia-toughened alumina/T i interfaces
is T iO [149]. The region in the oxygen chemical potential diagram (Fig. 7-1) where
this phase is stable corresponds to the oxygen chemical potential (indicated by the
arrow in Fig. 7-1) at which the decomposition reaction Al2 O3 2Al + 3O occurs.
In general, for zirconia/T i interfacial reaction processes involving active metal
brazing , it can be assumed [see 149] that:
This is indicated by a very negative heat of solution of oxygen in both bcc T i(O) and hcp
T i(O).
A rock-salt structure with oxygen and titanium atoms located in interpenetrating f cc lattices [34].
Provided there are no kinetic barriers to the precipitation of the thermodynamically stable
phase.
It is possible to make similar assertions with regard to other zirconia/T i interaction systems,
such as those described by Arroyave [51]
157
i) Because of the high mobility of solutes in liquid alloys , T i dissolved in the

brazing alloy is expected to diuse very rapidly towards the interface with zirconia, where it will react with the oxygen diusing out of the zirconia lattice.
To ensure charge neutrality, it is assumed that the proper charge transfer mechanism occurs at the interface and that the oxygen vacancies formed are always
charged-compensated.
ii) As oxygen diuses out of the zirconia lattice, unsaturated bcc T i(O) forms.
As oxygen continues to ow across the zirconia/T i interface a second phase,
hcpT i(O), precipitates and becomes increasingly saturated with oxygen until it
reaches the saturation composition with a stoichiometry corresponding to T i2 O.
If the oxygen chemical potential at the interface is xed at a suitable value, the
T iO1x phase can precipitate (see Fig. 7-2, horizontal arrow).
iii) When cooling the sample, solid-state transformations can take place, by which
the reaction layer is transformed to oxide phases stable (T i2 O3 , for example) at
low temperatures (see Fig. 7-2, vertical arrow) .
2000
Liquid
1800
BCC
1400
TiOx
HCP
1200
600
0.1
0.2
0.3 0.4 0.5
Mole Fraction X(O)
Ti2O3
0.6
0.7
Ti3O5
Ti4O7
TiO2
800
TiO
1000
Ti3O2
Temperature, 0C
1600
Figure 7-2: Calculated T i O phase diagram, calculated from the model by Lee [34]
According to Kirkarldy and Young [122], a typical value for the diusion coecient in liquid
metals is 1 109 m2 /s
T i2 O3 , for example.
158
7.3
Current Modelling of Coupled-Oxide Growth
As mentioned in the section above, due to the large gradient in the oxygen chemical potential across the zirconia/T i interface, a complex set of coupled phases are
likely to form. Multiple interfaces are thus created, whose temporal evolution is
determined by both kinetic and thermodynamic properties of the phases involved.
In order to understand the basic phenomena underlying these complex interfacial
reactions, increasingly-sophisticated diusion models have been developed over the
years [149]. As mentioned in Section 7.1, the present chapter deals with the development of a model that combines the solution of a Cahn-Hilliard( or, in modern terms,
Phase-Field )-type problem with assessed thermodynamic data obtained through
CALPHAD methods in order to model the coupled oxide growth of T i O phases
during zirconia/metal active brazing operations. Before the Phase-Field model is
developed, it is necessary to analyze the approaches currently used to model coupledoxide growth occurring during zirconia/metal joining processes.
The model developed by Torvund et al. [149] is a good example of current models
and it describes the behavior of zirconia-alumina ceramics, as they react with T i-rich
metals. In their model, Torvund et al. [149] considered that, because of the extremely
fast diusion rates in the liquid braze, there is always a T i-rich layer in contact with
the ceramic substrate. Once it is established that a T i layer is always in contact
with the zirconia surface, the process of interfacial layering was divided into three
independent stages:
i) Stage I. This stage corresponded to the rapid formation of unsaturated hcp
T i(O), followed by a gradual oxidation to T i2 O (see Fig. 7-2) through diusion
of oxygen. Note that this implied the complete transformation of bcc T i(O)
into hcp T i(O) due to fast oxygen diusion through the T i layer. In their
Ceramic Oxide
hcp-Ti(O)
Active Material
0.4
t3
x(O) 0.2
0
t2
t1
t3>t2>t1
distance
Figure 7-3: Stage I. of model by Torvund et al. [149].
model, Torvund et al. [149] considered that the diusion of oxygen through the
This is due to the large chemical anity between T i and O.

The model presented in this work does not include an explicit description of a variable indicating
the phase state of the system. However, the authors consider the model to belong in the Phase-Field
category since the phase state of the system is implicitly dened through its local composition.
T i2 O is the hcp T i(O) solid solution completely saturated with oxygen.
Since oxygen behaves as an interstitial atom in both the bcc T i(O) and hcp T i(O) phases,
it is expected that its diusion rate is very fast, compared to that of the oxide phases (i.e. T iO1x ).
159
hcp T i(O) was the rate controlling step. By setting a constant diusion coecient, they were able to obtain a mathematical expression for the concentration
of oxygen as a function of time in the form of a convergent series.
ii) Stage II. The second stage of the process considered that the already formed
T i2 O phase was further oxidized to T iO1x . In this case, the problem was solved
by considering s Stefan-like problem in which the velocity of the moving interface
(T i2 O/T iO1x ) is controlled by the mass conservation condition
JL = JR .
(7.2)
where JL and JR correspond to the ux entering through the left-hand side and
exiting through the right-hand side of the interface, respectively.
Ceramic Oxide
TiO(1-x)
Ti2O
Active Material
0.5
x(O) 0.25
distance
Figure 7-4: Stage II. of model by Torvund et al. [149].
iii) Stage III. This last stage involved the solid-state transformation of the hightemperature T iO1x phase into phases stable at lower temperatures, such as
T iO orT i2 O3 .
Despite the simplicity of the model described above and its ability of represent
the growth of a sequence of oxide layers in zirconia-T i interfacial reactions, there are
serious shortcomings that need to be addressed:
i) Due to the complexity of the diusion phenomena involved, the model described
above did not consider the diusion of oxygen into the high-temperature bcc
T i(O) phase, which is expected to be observed at the very early stages of the
zirconia/T i reactions.
ii) The model is unrealistic in that it separates the formation process of new phases
into arbitrarily dened stages. Consequently, dierent diusional models are applied depending on the particular stage of the process. A unifying transformation
mechanism cannot therefore be proposed for a more general analysis of such a
complex system.
160
iii) For the actual simulation of the growth process, it is necessary to rely on the
sharp-interface moving-boundary problem to represent all the moving interfaces
considered. From a computational point of view, the tracking of all the moving
interfaces requires the establishing of mass-conservation conditions across all interfaces. This is complicated when the transport properties of the phases dier
greatly: As has been noted by Lee and Oh [150] and others, the solution of
moving interface problems are usually accompanied by mass conservation errors
that need to be corrected using global mass balance conditions throughout the
entire system. As the number of simultaneously moving interfaces increases, the
uncertainty with regard to where to apply such corrections increases accordingly.
iv) Because of the complexities involved in the modelling of multiple moving-boundary
problems, the simulation of simultaneous diusional reactions taking place in
2 D systems, for example, is extremely expensive, from the computational
point of view
Given the limitations outlined above, it is necessary to develop a model capable
of describing the complex simultaneous diusional reactions occurring in multiple
moving boundary processes such as those occurring during ceramic/metal interfacial
reactions. In the following sections, it will be seen that by applying the Phase-Field
methodology, coupled with CALPHAD descriptions of the thermodynamic behavior
of the system it is possible to model, in a qualitative manner, these complex phenomena.
7.4
Thermodynamics and Kinetics of Heterogeneous Systems
In their paper on the thermodynamics of heterogeneous systems , Cahn and Hilliard

[147] established that the local free energy per molecule, f , in a region of space where
an intensive variable (such as composition) is non-uniform depends on both the local value of the variable and on the value that the variable takes in the surrounding
environment. As a concrete example, it is possible to consider a binary system, in
which the only variable that changes its value across space is composition. For this
particular example, the free energy of the system at any given point in space, as a
function of the local composition and the environment , will be given by:
f c,
c,
c, ( c)2 , . . . = f0 (c) + k0
c + k1
c + k2 ( c)2
(7.3)
where f is the free energy of the system (in J/mol) , c is the mole fraction of component B, f0 corresponds to the free energy as a function of concentration c in a
A heterogeneous system is dened as having a spatial variation in one or more of its intensive
scalar properties, such as composition, density, order parameter, and so forth.
The environment, in this sense, is given by the spatial derivatives of the concentration eld.
Note that f in theory corresponds to a constant-volume free energy, i.e. Helmholtz. However,
since in condensed phases the compressibility V is so small, the eect of pressure can be neglected
P
and the Gibbs free energy can be used instead.
161
homogeneous system and k0 , k1 , k2 are constants relating particular spatial derivatives of the concentration eld to the total energy of the system .
Eq. 7.3 indicates that in a heterogeneous system, the free energy of the system
will be a function of both the local value of the intensive variable and the spatial
variation of such variable. Since composition is a scalar material property, the free
energy of the system must be invariant with respect to its gradient , and therefore,
k0 = 0.
The total energy (F , in J) over a spatial volume V would be given by:
F =
f0 (c) + k1
c + k2 ( c)2 dV
(7.4)
where is the molar volume of the system .

By using the divergence theorem [see 147] we obtain:
(k1 c n) dS =
S
dk1
( c)2 + k1
dc
(k1 c) dV =
V
c dV
(7.5)
Since the external boundaries of the system enclosed within the volume V are of
no consequence , it is possible to make c n = 0 and therefore we have that:
dk1
( c)2 = k1
dc
f0 (c) + k2
dk1
dc
(7.6)
so Eq. 7.4 can be simplied into

F =
( c)2
(7.7)
In the general case, it can be safely assumed that dk1 = 0 . k2 is usually called
dc
the gradient energy coecient and relates the absolute value of the local gradient of
composition to the total value of the free energy of the system at any given point .
At regions of relative homogeneity, this parameter has no inuence whatsoever. At
In principle, k0 , k1 , k2 can be, in turn, functions of composition or other intensive variables, but
given the current lack of experimental information, they can be considered as constants.
A mirror symmetry operation performed on the system must not change the value of the energy
due to inhomogeneities in a scalar property.
Note that this requires that there is no change in molar volume with composition.
It is possible to assume that the system is so large that surface interactions with the outside
world are irrelevant.
Although this assumption is arbitrary, ignorance on the actual values for the gradient energy
coecient for real materials interfaces, makes it reasonable enough.
2
This parameter has units of Jm
mol
162
interfaces, however, where composition changes abruptly , the term containing k2 becomes dominant. The gradient energy coecient can therefore be conceptualized as
a proportionality constant that determines the energy that the system has to expend
in order to create an interface across which c has a nite value.
If the system is constrained to one dimension and it is further assumed that the
system is composed of two homogeneous phases joined through a diuse interfacial
area, a specic surface energy can be dened as [147]:
1
=
f (c) + k2
dc
dx
dx
(7.8)
where f (c) corresponds to the non-equilibrium free energy of material located at

the interface.
As the system approaches equilibrium, it follows that the surface energy must be
a minimum. By looking at the integrand in Eq. 7.8, it is obvious that the free energy
is minimized as dc 0, that is, as the interface becomes more diuse. In order to
dc
make the interface more diuse, it is necessary then to add more material into the
interface itself, increasing the f (c) term. A minimization problem thus results: by
2
using Eulers equation it is possible to nd that f (c) = k2 dc/dx , and nally
the interfacial energy of the system can be expressed as:
2
=
1
[k2 f (c)] /2 dx
(7.9)
An important result of this analysis is that, if the concentration or compositional

prole across the interface is the optimum so the surface energy of the system is minimized, the interface thickness is related to the gradient energy term, k2 or simply k,
through the relation k 1/2 .
It is possible to consider a system, whose energy is given by Eq. 7.7, which is
initially in a non-equilibrium state. In order for the system to reach equilibrium, its
local properties (i.e. c) must change. It can be assumed that the system will evolve
towards equilibrium through the fastest path possible and the kinetics of the system
In real interfaces, the change occurs within a few monolayers-width.

This surface or interfacial energy is the dierence, per unit area between the actual energy of
the system and that which it would have if the interface was neglected
If the integrand in Eq. 7.8 is identied as I, the equation I dc/dx I dc/

= 0 must
dx
be satised in order to make a minimum.
This assumption is based on pure variational arguments and it is assumed to be valid for any
163
can, using variational arguments, therefore be related to the rst variational derivative of its free energy [see 151] :
F =
f
2k
c
c c(x)
(7.10)
By using linear kinetic theory, it is possible to assume that the local ux of mass
occurring in any region of a non-uniform system evolving towards equilibrium will be
given by :
1
J = M F = M
f
2k
c
c(x)
(7.11)
where M is the atomic mobility given by:

M=
D
RT
(7.12)
with D being the diusion coecient, R, the universal gas constant and T the temperature.
Since c is a conserved parameter, the rate of change of c with respect to t, dc/dt
must be equal to minus the divergence of a ux, and therefore :
dc
=
dt
=M
f
2k
c
f
2k
c
c(x)
(7.13)
c(x)
Note that the rst term in Eq. 7.13 (Cahn-Hilliard equation) corresponds to Ficks
Second Law, with f corresponding to the chemical potential of c, (c). The second
c
term in Eq. 7.13 is related to the excess energy resulting from the creation of interfaces.
Originally, the Cahn-Hilliard equation was developed to describe the kinetics of
spinodal decomposition [148], which is a nucleation-free phase separation process occurring when a system, having a free energy vs. composition as described in Fig. 7-5,
has a composition within the spinodal (dened as the region enclosed by the condition
2G
< 0 ). Within the spinodal, the system is unstable with respect to any variation
c2
physical system.
For a general functional P (y) =
Q (y(x),
y) dV , the variational derivative of P [151, chap.
22] is given by Q Qy , which is basically Eulers Equation Collins [152, chap. 12]. According
y
to variational principles, when the variational derivative vanishes, y(x) is a extremal function.
2
mol
mol
Note that with M having units of m sJ , the ux, J, has the correct units of m2 s
1
Note that the factor must be dropped, as c represents local composition (non-dimensional),
and not concentration (moles per volume).
164
in composition and therefore it will minimize its energy by separating into two phases
(of the same crystalline structure) with compositions indicated in Fig. 7-5 by cA and
cB . In principle, the system will try to form as many interfaces as possible, however,
the surface energy that must be expended in order to create those interfaces will
limit the total interfacial area of the system.
1600
1400
GMR(FCC) J/mol
1200
1000
800
600
400
200
0
-200
cA
-400
-600
cB
0
0.2
0.4
0.6
x(Ag)
0.8
1.0
Figure 7-5: Free energy v.s composition diagram for the Ag Cu fcc system at 900 K.
7.5
Gibbs Energy Expression to model Coupled Growth System
The Cahn-Hilliard equation has been successfully used to describe complex processes
such as spinodal decomposition. It is therefore valid to inquire wether similar models can be used to describe other similarly complex phenomena, such as the coupled
growth of multiple reaction layers occurring due to interfacial reactions between a
ceramic and an active material. At rst glance, however, there is one issue regarding the implementation of Cahn-Hilliard -type models to describe diusional massive
transformations (such as the one described in Section 7.3) that needs to be considered: the issue in question regards the mathematical representation of the Gibbs free
energy of the system. In the case of spinodal decomposition, a single expression for
the free energy can be used to describe the thermodynamics of the system in its entirety, since it is comprised of a single phase . When attempting to model diusional
transformations in a system of several phases which can also vary in composition, the
(Gibbs) free energy of the system (at a given T and P ), as whole, must be necessarily
expressed as both a function of the local composition, c, and another, non-conserved
property, that reects the change in the local arrangement of the atoms belonging to
The area enclosed by the free energy curve and the common tangent linking the equilibrium
compositions cA and cB represents the non-equilibrium free-energy f (c) appearing in Eq. 7.8.
Which can exist at two dierent equilibrium compositions
165
dierent crystalline phases (see Fig. 7-6(a)):

GSystem = G (c, , T, P )
(7.14)
That is, for any given phase, , having a particular crystalline structure, it must
be possible to represent its Gibbs free energy as a function of composition. Furthermore, at any given composition, the Gibbs free energy of the system must also be
expressed as a function of an order parameter, , which is directly dependent on
the specic crystalline structure of the phase . For the system to undergo a phase
transformation (in Fig. 7-6(a), from bcc T i(O) to f cc T i(O), for example) at
a composition c, it is necessary to overcome an energy barrier, whose magnitude is
determined by the energetics of the atomic-rearrangements necessary for the transformation to take place. As can be seen in Fig. 7-6(a), the local minima along the axis
of the Gibbs free energy surface corresponds to the (metastable or unstable) equilibrium functions for the Gibbs free energy of each phase as a function of composition.
x 10
x 10
bcc
hcp
fcc
Energy Barrier for bcc->fcc

Transformation at x(O)=c
bcc
hcp
fcc
0. 5
GMR J/mol Ref: O2,g Ti,bcc
0. 5
GMR J/mol Ref: O 2,g Ti,bcc
1. 5
2. 5
1. 5
2. 5
3
3
3. 5
0
0.1
0.2
0.3
2
0.4
0.5
0.6 1
Order Parameter
(a) Gibbs free energy vs. xO and order

parameter .
3. 5
0.1
0.2
0.3
x(O)
0.4
0.5
0.6
(b) Gibbs free energy vs. xO
Figure 7-6: Gibbs free energy for the phases bccT i(O), hcpT i(O), f ccT i(O) (T iO1x )
at 12000 C, for the system T i O after assessment by Fischer [153].
In many applications of phase-eld models [e.g. 154] describing diusional transformations, the parameter is explicitly used to represent the phase-state of the
system at any given point in space. In these models, it is essential to nd the equilibrium compositions at both sides of the interface and the concentration prole within
the interface itself. In order to do this, it is necessary to perform Gibbs free energy
minimization calculations at every single point of the simulation domain. Once the
166
equilibrium compositions and the phase state of the system are dened, the timeevolution of the system is then determined by solving two types of coupled equations:
the Cahn-Hilliard equation for the conserved variables and the Allen-Cahn equation
(for a more detailed explanation please refer to Allen and Cahn [155].) for the nonconserved parameter. In addition to providing the proper thermodynamic models
for each of the phases involved and realistic values for the atomic mobilities of the
components comprising the system, mobilities for each of the interfaces involved in
the simulation must be provided. If the system is known not to be limited by surface
attachment kinetics, the interface mobilities must be chosen in such a way that the
time-evolution of the phase order parameter, , does not aect the kinetics of the
phase transformation.
Although there is a physical argument for the existence of an order parameter
, most of the current phase-eld models use arbitrary mathematical expressions to
describe the Gibbs energy dependence on the value of this parameter. If the system
to be modelled is known to be controlled by bulk-diusion mechanisms, rather than
by surface attachment limited kinetics, it may be possible to avoid the use of the
parameter (and therefore of Allen-Cahn-type equations describing the kinetics of
non-conserved variables) and instead consider the phase-state of the system in an
implicit manner. In this case, the concentration, c, acts as an indicator of both the
thermodynamic and phase-state of the system:
GSystem = G (c, T, P )
(7.15)
This is equivalent to collapsing the Gibbs free energy surface in Fig. 7-6(a) across
the axis, as shown in Fig. 7-6(b). In order to justify this procedure, it can be supposed that the energy barrier for a phase transformation (at a given composition c)
is small enough so it can be neglected. Thus, the system will always be in a state for
which the local Gibbs energy is a global minimum at a given composition c, changing
its crystalline structure (i.e. the value of the parameter ). In a mathematical sense,
this last assertion is equivalent to assuming that the Gibbs energy of the whole system will be given by the envelope of the Gibbs free energy functions for the dierent
phases comprising it, along the axis projection.
GSystem = min GbccT i(O) , GhcpT i(O) , Gf ccT i(O) {xO }
(7.16)
Fig. 7-7 shows the Gibbs free energy of the system bcc hcp f cc in the T i O
binary after the operation in Eq. 7.16 was performed on the projection of the Gibbs
free energy of the phases bcc T i(O), hcp T i(O) and f cc T i(O) along the axis,
as shown in Fig. 7-6(b). The insert in Fig. 7-7 illustrates a schematics of the process.
Note that, in this operation, the points at which the common tangent (equilibrium
between phases and ) touches the Gibbs free energy curve remain the same, that
is, the equilibrium compositions of the two-phase eld do not change. Since the slope
of the common tangent is also the same, the chemical potential at which both phases
are at equilibrium remains invariant. The shaded area in Fig. 7-7, between the com167
x 10
bcc Ti
0.5
2,g
2.5
GMR
System
J/mol Ref: O
1.5
3.5
0.1
0.2
0.3
x(O)
0.4
0.5
0.6
0.7
Figure 7-7: min GbccT i(O) , GhcpT i(O) , f cc T i(O) at 12000 C.

2. 5
x 10
Excess Contribution from

the Non homogeneous
Chemical Potential
(O) Ref:O
2,g
3. 5
4. 5
5. 5
bcc
6
fcc
hcp
0.1
0.2
0.3
x(O)
0.4
0.5
0.6
0.7
Figure 7-8: Oxygen chemical potential (O) as a function of mole fraction of oxygen xO .
mon tangent and the envelope of the Gibbs free energy, corresponds to the excess
energy of the material within the interface, or f (c) in Eq. 7.8. Note that f in
c
Eq. 7.13 can be considered to be equivalent to a chemical potential .
The oxygen chemical potential, (O) depicted in Fig. 7-8 was obtained through
the numerical tting of the function represented by Fig. 7-7 and it is therefore an
G
ni
n ,T,P
The oxygen chemical potential was represented through a polynomial of order N + 1. P =

1 108 [1.9124, 2.5315, 1.1574, 0.2117, 0.0164, 0.0058] represents the vector with the coecients
of the polynomial given by (c) = P (1) cN + + P (N + 1) , where c corresponds to the oxygen
168
approximation to the real values obtained through the use of the thermodynamic
models for the bcc T i(O), hcp T i(O) and f cc T i(O) phases as developed by Fischer [153] . Comparison with the oxygen chemical potential plotted in Fig. 7-1, the
agreement is good.
7.6
1D Simulation Results
In order to simulate the coupled oxide growth process described in Section 7.3, the
Cahn-Hilliard equation (Eq. 7.13 in Section 7.4) was solved using the numerical methods briey described in Appendix E. The thermodynamic properties for the T i O
phases were obtained from the CALPHAD model developed by Fischer [153] and
were later modied (see Section 7.5) so the Gibbs free energy of the system could be
described with a single mathematical expression. The boundary conditions for all the
simulations described in this section were: xed chemical potential on the left-hand
side of the simulation domain, and closed boundary (i.e. zero ux) on the right-hand
side.
In order to analyze the ability of Phase-Field methods to describe complex multiple moving boundary diusion problems, a series of numerical experiments were performed using the real thermodynamic descriptions and somewhat simplied kinetic
parameters. For the simulations presented in this section, the mobility appearing in
Eq. 7.13 was given by :
M
D
=
(7.17)
RT
Note these simulations, mobility, M is a constant and is the same for the three phases
involved . Although this assumption may hold for the cases of the bcc T i(O) and
hcp T i(O), it is not clear that the assumption is valid for the f cc T i(O). This
simplication of the model can be easily removed by allowing the mobility to be a
function of composition, but for the purposes of illustrating the applicability of PhaseField methods in these kind of diusional problems, its removal is not essential.
As is noted in Appendix E, the gradient energy coecient k in the Cahn-Hilliard
equation is a fundamental physical parameter of real interfaces whose value determines
their width and basic properties. In principle, the smaller the value for k, the smaller
the actual with of the interface, since, as noted in Section 7.4,
k 1/2
(7.18)
mole fraction, or xO .
Note that in principle, (O) must approach as xO 0
bcc T i(O), hcp T i(O) and f cc T i(O).
The entire compositional range examined.
This basically accelerates the simulation relative to the dimensions of the system.
For the purposes of this preliminary analysis of this Phase-Field model, D was set to have a
value of 10.
169
Given that natural interfaces have a with of just a few monolayers, k would have
to have a very small value. In numerical simulations, however, it is not currently
possible to simulate systems with such small interface widths, and therefore, the
following relationship (see Appendix E) was used:
3
h
2
k=K
(7.19)
where h is the width of the computational grid (in units of length), while K is a
proportionality constant that is adjustable.
7.6.1
Numerical Simulations: Inuence of k Parameter
Concentration Profile K=1E+02
0.55
Concentration Profile K=1E+04
0.55
t =4.7744
t =4.7744
0.5
0.5
0.45
O %at
O %at
0.45
0.4
0.4
0.35
0.35
0.3
15
16
17
18
19
20
Distance
21
22
23
24
25
0.3
15
(a) Concentration Prole. nx = 100,

K = 1E + 02
16
17
18
19
20
Distance
21
22
23
24
25
(b) Concentration Prole. nx = 100,

K = 1E + 04
Figure 7-9: Eect of the value of k on the width of interfaces created trough the evolution
of the Cahn-Hilliard equation.
As noted above in this section, the value assigned to the parameter k has a great
inuence on the properties of the interfaces formed during diusional transformations
governed by the Cahn-Hilliard equation. In principle, by increasing the value of k, the
interfaces in the system would become increasingly diuse, as can be seen in Fig. 7-9.
Fig. 7-9(a) shows the calculation results for a system for which the adjustable constant, K (see Eq. 7.19) has been set with a value of 1 102 . Note that the width of
the f cc/hcp interface is noticeably less diuse than that shown in Fig. 7-9(b), which
was obtained through the evolution of the Cahn-Hilliard equation using K = 1 104 .
This discrepancy can be reconciled by considering that a diuse interface, by its very
nature, is impossible to dene in a precise manner .
In principle, a diuse interface extends from to +
170
Besides its eect on the interface width, k has another, rather unexpected, eect
on the behavior of systems evolving under the kinetics described by the Cahn-Hilliard
equation: the value of k has a noticeable eect on the global mass conservation behavior of the numerical simulation.
In order to calculate the total mass in the system, at any given time, it is possible
to integrate, over time, the ux of oxygen entering the system through its left (xed)
boundary:
t
J(t)x=0
M ass (t) =
(7.20)
Additionally, one can calculate the total mass in the system at any given time t, by
integrating the concentration prole over the entire dimension of the computational
domain:
x=L
M ass (t) =
Conc.(x)t dx
(7.21)
x=0
K=1E+02
K=1E+03
K=1E+04
K=1E+05
% Error Mass
2.5
1.5
0.5
0
100
150
200
250
300
350
400
# Inner Nodes
450
500
550
600
Figure 7-10: Global mass error as a function of number of inner nodes.
In principle, the results obtained from Eq. 7.20 and Eq. 7.21 should not dier.
However, as can be seen in Fig. 7-11, k has a marked eect on the global mass conservation error present during the numerical simulations. For relatively small values
for the adjustable parameter K (K < 1 104 ), the global mass error lies within
acceptable ranges. The discrepancies within this range are mostly determined by the
numerical approximations utilized in the discretization of the C H equation, the
numerical integration of the concentration prole, and the numerical approximation
171
of the local ux at the left-hand boundary of the simulation domain, as shown in

Fig. 7-10, where the global mass error in the numerical simulations decreases as the
number of grid points in the computational domain increases . Beyond K = 1 105 ,
however, the global mass error increases in a rapid manner, becoming unacceptable
at values of K 1 105 . As of now, there is no clear explanation for this discrepancy. However, it may be linked to the fact that, as k increases, the interface
becomes increasingly diuse (see Fig. 7-9) and more material is transferred into the
non-equilibrium interfacial region. This result may be of fundamental importance
when using Phase-Field models to describe phase transformations in systems with
large interfacial areas, compared to their volume (or total area, in the case of 2-D
simulations).
3
nx100
nx200
nx 300
nx400
nx 500
nx 600
% Error Mass
2.5
1.5
0.5
0
2
2.5
3.5
4.5
Log(K)
Figure 7-11: Eect of K on the global mass error of C-H numerical simulations of coupled
oxide growth. Dierent numbers of inner grid points are examined.
7.6.2
Analysis of Numerical Results
In Fig. 7-12, the concentration and chemical potential proles at a given time t are
presented. From this gure, two important results are worth mentioning:
i) Despite the relative shallowness of the interfacial excess energy hump of the
Gibbs free energy function of the T i O system , the equilibrium interfacial
compositions are well dened in the plot . According to the gure, the bcc
T i(O) phase is already saturated with oxygen at this stage of the simulation so
That is, as the numerical approximations become increasingly precise.

Refer to the gure caption for simulation parameters.
This hump can be considered to correspond to a non-equilibrium region.
Compare Fig. 7-5 to Fig. 7-7.
This is not surprising, since it is always possible to nd a linear transformation of the analytical
description of the Gibbs energy of the system that can amplify its non-equilibrium regions.
172
Concentration Profile
0.6
-3.0
t =4.7744
x 10
Chemical Potential Profile

t =4.7744
-3.5
0.5
-4.0
0.4
mO
O %at
-4.5
0.3
0.2
fcc
-5.5
fcc
hcp
hcp
-6.0
0.1
-5.0
bcc
-6.5
bcc
0
20
40
60
Distance
80
100
20
40
60
Distance
80
100
Figure 7-12: Concentration and chemical potential proles. Phase-Field simulation of coupled oxide growth at T = 1573. The K factor in Eq. 7.19 has been set at K = 100. The
number of inner grid points is 600.
there is only a hcp bcc phase transformation occurring in those regions of the
simulation domain.
ii) The system, therefore, nds the correct equilibrium interfacial compositions
by itself without the need for performing Gibbs free energy minimizations at
each point of the computational domain. Solving the Cahn-Hilliard equation
eectively minimizes the global energy of the system.
iii) It is worth noting that the calculated prole for the oxygen chemical potential,
(O), is continuous along the length of the simulation domain. This is consistent
with the fact that the system is always at local equilibrium, even in the nonequilibrium regions (from the global perspective) at the interfaces. This local
equilibrium approximation is also used in simulation methods that rely on sharp
interface models, such as the one described by Lee and Oh [150].
As noted in 7.6.1, the value of k greatly inuence the mass conservation properties of numerical integrations of the Cahn-Hilliard equation. However, with the
proper selection of the gradient energy coecient term, global mass conservation discrepancies lie within reasonable ranges, considering the discretization of the model.
Fig. 7-13 shows that, for values of K smaller than 1, 000, the general condition:
173
Total Mass in the System vs. Time
Relative Error in Mass vs. Time
x 10-3
Flux integratio n
Profile Integratio n
50
45
2.5
40
35
Mass
Rel. Error
30
25
1.5
20
15
10
5
0
0.5
0
10
15
20
Simulation Time, sec
25
10
15
20
Simulation Time, sec
25
Figure 7-13: Global mass conservation for the solution of the Cahn-Hilliard equation. 600
grid points and K = 100.
t
x=L
J(t)x=0 dt =
0
Conc.(x)t
(7.22)
x=0
is satised. Moreover, it is also evident (right graph, in Fig. 7-13), that the discrepancy between the
ux integration and prole integration calculations to determine the total mass
of the system decreases exponentially as time progresses. Initial transients as the
system starts to evolve, are therefore the greatest factor in determining conservation
inconsistencies.
It can be assumed that at the high temperatures (T 1, 0000 C) at which these
diusional processes take place, coupled oxide growth in the T i O system is a diffusion limited problem . Therefore, it is to be expected that the displacement of the
interfaces is governed by the square root law:
x
(7.23)
The mobility of the interface constitutes no limiting factor whatsoever on the behavior of the
system.
174
Displacement vs. time
100
hcp/bcc
fcc/hcp
80
80
70
70
60
60
50
hcp/bcc
fcc/hcp
90
50
90
Displacement vs. time 1/2
100
40
40
30
30
20
20
10
10
10
15
20
25
t 1/2
Figure 7-14: Displacement of interfaces versus time and time1/2 .
Fig. 7-14 shows that the displacement of the bcc/hcp interface is faster than that
of the f cc/hcp interface. Unnam et al. [156] reported that generally, the bcc/hcp
rate of transformation is much faster than the one involving a higher oxide, such as
T iO1x (as in this case) or T iO2 , as reported in [156]. As can be seen in Fig. 7-14,
at very short times after the simulation commences, both interfaces seem to obey the
square root law. As soon as the fast-moving hcp/bcc interface disappears (see square
in Fig. 7-14), the slower interface accelerates (transition point indicated by a circle in
Fig. 7-14).
As long as the bcc T i(O) phase is present in the system, the equilibrium concentrations at the hcp/bcc and f cc/hcp interfaces are xed. The ux of oxygen atoms
across the hcp T i(O) phase (from the f cc T i(O) towards the bcc T i(O)) is
controlled by the gradient in the chemical potential within this phase, which in turn
is given by the dierence in chemical potential at both interfaces, (O) , divided
by the thickness of the hcp T i(O) layer,h:
J
(O)
h
I.e., the bcc T i(O) phase has completely transformed into hcp T i(O).
175
(7.24)
Since the dierence in the equilibrium in chemical potentials stays constant (as
long as both interfaces are present), the oxygen ux within the hcp T i(O) layer is
controlled by its instantaneous thickness, which increases as the square root of time.
The ux will therefore tend to decrease over time until the bcc T i(O) layer disappears. At this stage, the hcp T i(O) layer stops growing and begins saturating with
oxygen. It is at this point when the f cc hcp phase transformation begins to accelerate, as can be seen in Fig. 7-14. The change in the rate of this phase transformation
is due to the fact the hcp T i(O) layer saturates rapidly with oxygen, in the absence
of the bcc T i(O) phase. As the saturation progresses, any additional amount of oxygen entering this layer leads to its immediate transformation into f cc T i(O). From
an alternative point of view, it could be said that the hcp/bcc interface exerts a back
pressure over the slow-moving f cc/hcp interface, as soon as the hcp/bcc interface disappears, the back pressure is relieved and the displacement of the remaining f cc/hcp
interface accelerates. It is thus evident that the growth of the three individual phases
considered in this system is coupled through the thermodynamic properties of the
interfaces.
Although the system conserves mass and behaves as if it were controlled by
volume-diusion processes (as indicated by the x vs. t1/2 plot in Fig. 7-14), it is
worth examining its behavior with respect to the Stefan Condition , as should be
applied at each of the interfaces:
dX (t)
(7.25)
dt
where X (t) represents the position of the interface at any given time, t, cL and cR
represent the equilibrium compositions at the left and right-hand sides of the interface, and J(t)in , J(t)out are the inward and outward mass uxes evaluated at the
interface.
J(t)in J(t)out |X(t) = (cL cR )
Eq. 7.25 basically states the local condition for mass conservation at any interface,
X. Since the global mass conservation condition (depicted in Fig. 7-13) holds, this
local condition must also hold, and it should be possible to test it by analyzing the
local behavior of the interfaces resulting from the simulation. Due to the very nature
of a diuse interface (no sharp boundaries divide the phases), a rigorous test of the
Stefan Condition is impossible unless one integrates the uxes over an innite range,
which is basically what the plot of Fig. 7-13 does.
Fig. 7-15 presents, in a very crude way, a verication of the Stefan Condition.
Note that although it is not possible to precisely determine the dierence in equilibrium compositions, the product of such approximate dierences and the interfacial
velocities at that time roughly correspond to the equally approximate dierence in
the uxes coming in and out of both interfaces. A larger dierence in the uxes at
In the limit where the diuse interface approaches a sharp one.
176
0.6
Concentration Profile
Flux
t =4.7748
t =4.7748
2.2
J~0.4
0.5
c~0.18
1.2
Local Flux
O %at
v~7.5
1.6
1.4
0.3
hcp/bcc
fcc/hcp
1.8
0.4
Velocity vs. time
10
0.8
0.2
3
0.6
2
0.4
c~0.15 J~1.2
0.1
50
Distance
100
v~2
0.2
50
Distance
100
10
Figure 7-15: Stefan condition at the hcp/bcc and f cc/hcp interfaces.
the bcc/hcp interface naturally leads to a greater interface displacement rate.

Fig. 7-16 shows the concentration proles at dierent times for the simulation
of coupled-oxide growth using the kinetics described by the Cahn-Hilliard equation.
Note that, due to fast interstitial diusion, the bcc T i(O) phase reaches the saturation composition in a fast manner. As this saturation progresses, the hcp/bcc and
f cc/hcp interfaces keep advancing at dierent rates, as indicated in Fig. 7-14. Note
that, due to local equilibrium considerations, the oxygen mole fraction at both ends of
the eld corresponding to the hcp phase remain xed. Consequently, the gradient in
composition (And chemical potential) decreases as the phase eld expands, decreasing the transformation rates and characteristic evolution rate of the system as a whole.
Fig. 7-17 shows the calculated proles for the oxygen chemical potential at different times during the simulation. It is evident that, according to the principle of
local equilibrium, the prole for the oxygen chemical potential constitutes a smooth,
monotonically-decreasing function of distance and the system evolves down the gradient in the oxygen chemical potential until it reaches an equilibrium at which further
minimization of the Gibbs free energy is not possible.
It reaches the bcc T i(O) equilibrium composition in the bcc + hcp two-phase eld, according
to Fig. 7-2.
Reference State: O2,g
177
0.6
0.5
0.4
0.4
% at O
% at O
0.5
0.3
0.2
0.2
t =2.4624
t =4.9549
t =7.4474
0.1
0
0.3
20
40
t =9.9399
t =12.4324
t =14.9249
0.1
60
80
100
20
Distance
60
80
100
80
100
Distance
0.6
0.5
0.5
0.4
0.4
% at O
0.6
% at O
40
0.3
0.2
0.2
t =17.4174
t =19.9099
t =22.4024
0.1
0
0.3
20
40
t =24.8949
t =27.3874
t =29.8799
0.1
60
80
100
Distance
20
40
60
Distance
Figure 7-16: Concentration proles at dierent times.
7.7
2D Simulation Results
In Section 7.6, it has been shown that it is possible to model the coupled growth of T i
O phases using CALPHAD descriptions of their thermodynamic behavior, together
with a diuse interface description of the kinetics of the phase transformation. As
has been demonstrated, the displacement of the multiple interfaces follows the squareroot law, which is to be expected if the phase transformation processes are diusioncontrolled. It has also been shown that the model satises global mass conservation
conditions. Moreover, it has been demonstrated, in a somewhat crude manner, that
the Stefan condition is satised at each of the interfaces.
Given that the present model is physically-sound , it is interesting to explore its
behavior at higher dimensions. To this eect, a 2 D simulation of the coupled oxide
growth model described in Section 7.6 has been implemented. Fig. 7-18 presents a
2 D simulation of the coupled oxide growth in the T i O system, where the only
phases considered are bcc, hcp and the f cc T i O oxide solid solutions.
The simulation represented by Fig. 7-18 uses, as initial condition, a uniform composition on the bottom boundary of the computational domain, corresponding to the
f cc T i(O) phase eld. Because of this initial condition, the behavior of the system
Essentially, the Stefan condition is a local mass-conservation condition for each of the moving
interfaces.
It is important to note that realistic values for the atomic mobilities of oxygen in each of the
phases have not been incorporated yet in the model. A concentration-dependent mobility eld can,
however, be incorporated to the model, provided reliable data for each of the phases is available.
178
x 10
x 10
3. 5
3.5
4
4.5
(O)
(O)
4. 5
5
5. 5
5
5.5
t =2.4624
t =4.9549
t =7.4474
6
6. 5
0
20
40
6.5
60
Distance
80
100
20
40
60
80
100
60
80
100
Distance
x 10
x 10
3.5
4.5
4.5
(O)
3.5
(O)
t =9.9399
t =12.4324
t =14.9249
5
5.5
5
5.5
6
6.5
0
t =17.4174
t =19.9099
t =22.4024
20
40
6
6.5
60
Distance
80
100
t =24.8949
t =27.3874
t =29.8799
20
40
Distance
Figure 7-17: Chemical potential proles at dierent times for coupled-oxide growth simulation using the Cahn-Hilliard formalism.
(a) Concentration prole in 2-D simulation.
(b) Chemical potential prole in 2-D

simulation.
Figure 7-18: 2-D simulation of coupled oxide growth. Planar initial conditions.
is basically the same as the one that has been observed in the 1 D simulations
presented in the previous section. It is perhaps more interesting to analyze the behavior of the system when the initial condition corresponds to a perturbation in the
concentration eld.
Fig. 7-19 shows a 2D calculation of the coupled growth of multiple T iO layers
179
Figure 7-19: 2-D simulation. Perturbations in the concentration prole were incorporated
as initial conditions.
considering, as initial condition, a semi-sinusoidal region with an initial composition

corresponding to the f cc T i(O) phase. As seen in Section 7.6, the growth of the
f cc T i(O) phase is much slower , than the growth of the hcp T i(O) phase. In
this 2 D simulation, therefore, it is to be expected that diusion across the initial
f cc T i(O) region is so slow compared to the diusion across the bcc T i(O) and
hcp T i(O) regions, that any perturbation in the concentration prole would tend
to disappear over time. According to this simulation, however, in addition to the
much slower f cc T i(O) hcp T i(O) transformation, it is observed that in fact
the initial f cc T i(O)/bcc T i(O) interfaces actually recedes, since it is thermodynamically unstable with respect to an hcp T i(O)/bcc T i(O) interface. It is seen
then that, according to this model, a sequence of planar reaction layers is the most
stable interfacial conguration. In the next section it will be shown that experiments,
reported elsewhere, conrm this nding.
7.8
Experimental Verication of Phase-Field Model
Although there is no experimental work reproducing the exact conditions that were
simulated in this chapter, it is still possible to establish the validity of the assump
Due to a lower diusion coecient
180
tions made while developing the model, as well as the main ndings regarding the
qualitative behavior of the coupled layer growth simulation. To this eect, in this
last section several of the key aspects of the model described in Section 7.7 through
Section 7.7 will be compared against experimental facts obtained elsewhere.
7.8.1
Examining the Local Equilibrium Hypothesis
The assumption of local thermodynamic equilibrium is one of the key hypothesis regarding both diuse and sharp interface phase transformation models. This implies
that there are no abrupt changes in the chemical potentials of the systems components. In diusional transformations, this assumption of local equilibrium implies
that the chemical composition at both sides of the interface corresponds to the equilibrium compositions given by the equilibrium phase diagram. Although the local
equilibrium assumption seems to be acceptable apriori, there are some experimental
results that question its validity [156].
In his early experiments on the oxidation of titanium, Jenkins [157] found that
the oxygen content in the hcp T i(O) side of the hcp T i(O)/oxide interface did not
reach the saturation limit ( 30 at.%, see Fig. 4-11(b)) that would be expected if the
hcp T i(O)/oxide interface were at equilibrium. In those experiments, it was found
that, after holding the samples for 72 hr., the concentration at the hcp T i(O)/oxide
interface was 12 at.% at 6500 C, 19 at.% at 8000 C, and 25 at.% at 9000 C. In later
oxidation experiments (650 7000 C), Hurlen [158] found that the initial oxygen
content of the hcp T i(O) phase in contact with the oxide layer was within the
14 15 at.%.After analyzing the samples, Hurlen [158] concluded that the composition at the ceramic/metal interface corresponded to that of the T i6 O . In a much
more recent series of experiments, Unnam et al. [156] found that the saturation
concentration of the hcp T i(O) phase, in equilibrium with oxide phases, was not
reached at low temperatures (5900 C). However, at temperatures above 8000 C the
hcp T i(O) saturation composition ( 30 at.%) was reached in very short times.
It is important to note that the fact that non-equilibrium compositions are found
at the interfaces corresponding to low-temperature oxidation may be explained by
low interface mobilities Martin et al. [159, p. 80]. In this case, the model develop in
this chapter could not be applied, unless it is modied to incorporate the phase state
of the system in an implicit manner.
In a very recent experiment on the early stages of laser-induced oxidation of titanium, Lavisse et al. [160] veried the local equilibrium condition for T i O phases
through XRD, GIXRD, XPS and EDX analyses of their samples. At least at the
early stages of their oxidation experiments, Lavisse et al. [160] were able to identify several hexagonal compounds with compositions in agreement with the T i6 O,
This T i6 O is an hcp-based phase in which oxygen and vacancy interstitials are ordered in
alternating sequences along the perpendicular axis to the basal plane [58].
181
T i3 O and T i2 O stoichiometries. The last stoichiometry, T i2 O, corresponds to the

hcp T i(O) solution, fully saturated with oxygen interstitials. In some of their samples, the authors were able to identify (using GIXRD) the T iOx .
Even if the local-equilibrium condition were not to be reached at low temperatures, the evidence available in the literature appears to conrm the local equilibrium
hypothesis for diusional transformations in the T i O system, at least for temperatures above 7500 C. Since the simulations performed in this chapter at at 1000 C,
the local-equilibrium implications of the Cahn-Hilliard model seem to be acceptable.
7.8.2
Experimental Verication of Diusion-Controlled Kinetics
As has been shown in Section 7.6, the Phase-Field model developed in this chapter
produces a sequence of dierent T i O phases, whose time evolution is controlled by
volume diusion . According to this model, there are no surface-attachment limiting
eects occurring at the interface. This behavior seems to be conrmed by a number
of experimental works on the oxidation of T i-based metals:
In their work on the low-temperature oxidation in titanium, Rogers et al. [161]
explored the diusion of oxygen in T i within the 127 5270 C range, for periods ranging from 1 to 100 hr. under several pressure conditions. Oxygen concentration proles
were measured using Auger spectrography. From their analysis on the behavior of
the oxygen concentration prole, as a function of time, Rogers et al. [161] observed
that the sequence of T i O compounds (including T iO2 , T iO, T i2 O, and so forth)
appeared to grow following the square-root-of-time law, and therefore were diusion
controlled.
In a higher-temperature series of experiments, Unnam et al. [156] studied the
oxidation kinetics of T i sheets within the 593 7600 C range. The thickness of the
oxide resulting from the oxidation was measured by optical means, while the oxygen concentration in the hcp T i(O) layer was determined using previously calibrated micro-hardness measurements. According to their measurements, the solidstate phase transformation processes occurring in all their samples were diusioncontrolled, since all the moving interfaces that were observed in the samples followed
the square-root-of-time law.
As has been noted in Section 7.4 and Section 7.5, in cases when the kinetics
of the phase transformations are not limited by interface mobility eects, the CahnHilliard equation should be capable of describing the relevant kinetic behavior of such
transformations. Given the evidence presented above, it is reasonable to expect that
the coupled layer growth in the T i O system at even higher temperatures will be
This phase is an f cc T i(O) solid solution, with a structure identical to the N aCl prototype.
The displacement of the dierent interfaces is governed by the square-root-of-time law, X (t)
t.
182
controlled by volume diusion. Therefore, the use of the Cahn-Hilliard equation for
this particular problem seems to be justied.
7.8.3
Experimental Evidence for the Stability of Planar vs. Undulated

Interfaces in the T i O System
In Section 7.7, it was shown that, according to the Cahn-Hilliard model described
in this chapter, a planar sequence of layered T i O phases is more stable than an
undulated-type of interface. This is to be expected if, during the phase transformations, any possible capillary eects due to the local curvature of the interfaces
created, were suppressed by the dissipative nature of the diusion-controlled kinetics
described by the model.
Although there has been a considerable amount of experimental work on the
oxidation behavior of titanium and titanium alloys [156158, 160, 161], there has only
been, to the best of this authors knowledge, one report on the behavior of multiple
moving boundaries as a result of oxygen transport across T i O phases. In their
work on the oxidation of commercially pure titanium, Unnam et al. [156] subjected
several titanium samples to an air atmosphere at dierent temperatures (see 7.8.2).
In their experiments, Unnam et al. [156] were able to observe the formation of a
titanium oxide (T iO2 , or rutile), as well as the formation of a T i O solid solution,
corresponding to the hcpT i(O) phase. Within the hcpT i(O) phase they were also
able to determine, through metallographic etching techniques, a moving boundary,
parallel to the oxide-metal interface. From their observations, Unnam et al. [156]
were able to determine the following:
i) The moving boundary represented a demarcation between relatively oxygen-rich
and oxygen-poor regions of the hcp T i(O) matrix. From their micro-hardness
measurements , however, they were not able to determine any discontinuity in
the oxygen concentration across the interface.
ii) The displacement of this moving boundary, X(t), had a denite correlation with
time and temperature, and its displacement obeyed the square-root-of-time law.
iii) The composition of this moving boundary was found to be, according to the
micro-hardness measurements, around 5 at.%.
Since the oxidation experiments by Unnam et al. [156] were performed below the temperature (t < 8000 C) at which the bcc T i(O) phase becomes stable
( 8900 C), the formation of the moving boundary could not be attributed to the
hcp T i(O) bcc T i(O) transformation that should occur at high temperatures.
In principle, it may be possible that this moving boundary resulted from the formation
As has been noted in 7.8.2, Unnam et al. [156] used calibrated micro-hardness measurements
on the hcp T i(O) phase in order to determine its oxygen content.
183
of an hcp-based ordered interstitial solution [58]. However, the measured oxygen concentration ( 5 at.%) does not correspond to any reported T iO interstitial solution.
Despite the fact that the experimental conditions in the experiments by Unnam
et al. [156] do not correspond to the conditions of the simulations presented in Section 7.6 and Section 7.7 , it is perhaps interesting to observe the interfacial structure
resulting from these experiments. Fig. 7-20 shows a typical result from the titanium
oxidation experiments by Unnam et al. [156]. As can be seen, the interphase morphology is planar and there are two clearly identiable interfaces within the T i O
region. As has been noted above, the experimental conditions that produce the interfacial morphology in Fig. 7-20 do not correspond to the simulations presented in
Section 7.6 and Section 7.7. However, the similarity between Fig. 7-20 and Fig. 718(a) is remarkable.
Figure 7-20: SEM micrograph of a T iA55 sample oxidized at 6490 C for 65.9hr. A=matrix,
B=moving boundary, C=oxide and D=copper coating. Originally, Fig. 7 in Unnam et al.
[156, page 242]. Reproduced with permission from the publisher (Kluwer Academic Publishers).
Metin and Inal [162] studied the layer growth during titanium ion-nitridation
in nitrogen plasma, at dierent nitrogen partial pressures within the 800 10800 C
range. In their work, the authors found that the layers formed corresponded to what
could be expected from the T i N phase diagram. As can be seen in Fig. 7-21, the
T i N phase diagram closely resembles the T i O phase diagram (see Fig. 4-11(b)),
at compositions of nitrogen and oxygen below 50 at.%. Even more signicantly, in
the T i N solid solutions, N behaves as an interstitial and the crystal structures of
The oxidation experiments by Unnam et al. [156] were performed under air, so the formation
of the rutile, T iO2 oxide was thermodynamically favorable, compared to that of lower oxides, such
as T iO or T i2 O3 .
184
these phases [163] have a one-to-one correspondence with the respective T iO phases
(bccT i(O) bccT i(N ), hcpT i(O) hcpT i(N ), f ccT i(O) f ccT i(N )).
3500
Liquid
2500
2000
TiN
1500
bcc
1000
500
hcp
0.1
Ti2N
Tmperature 0C
3000
0.2
0.3
0.4
Mole Fraction x(N)
0.5
0.6
Figure 7-21: T i N phase diagram, calculated from the thermodynamic model by Othani
and Hillert [164].
In Fig. 7-22, an optical micrograph of one of the ion-nitridation experiments

by Metin and Inal [162] is presented. This nitridation experiment was performed
at 9000 C. As can be seen by examining the T i N phase diagram in Fig. 7-21,
the layering sequence (T iN/T i2 N/hcp/bcc) observed, corresponds to what could be
expected from a series of coupled phase transformations in the T iN system, along a
N chemical potential gradient. By comparing Fig. 7-22 to Fig. 7-18(a), it is possible
to identify several similarities between the calculation for the T i O system and the
titanium nitridation experimental results reported by Metin and Inal [162]:
i) The layering sequence corresponds, in both cases, to what should be expected if
the diusional transformations occurred along a chemical potential gradient of
the interstitial element (oxygen or nitrogen).
ii) Excluding the T i2 N nitride, the rate of the f cc hcp transformation is much
slower than that of the hcp bcc transformation .
In all the solid solution phases in both the T i O and T i N systems, the diusion of the
interstitial is expected to be fast, with similar atomic mobilities for all the phases. However, the
site occupancy of both N and O interstitials, relative to interstitial vacancies in the f cc phase is
signicantly less that that of the bcc and even the hcp phases.
185
iii) A planar interfacial morphology seems to be more stable than an undulated one.
Thus, capillary eects, due to local curvature of the dierent moving interfaces do
not play an important role on the morphology of the reaction product layering.
Figure 7-22: Optical micrograph of the cross-sections of the ion-nitrided titanium at 9000 C.
Original gure appeared as Fig. 4 a) in Metin and Inal [162, page 1823]. Reprinted with
permission from the publishers (TMS).
As can be seen from Fig. 7-22, the stability of planar interfaces with respect to
undulated ones in these types of phase transformations seems to be justied from the
experimental point of view.
7.9
Sensitivity of the Model To the Thermodynamic Description
Before closing this chapter, it is necessary to consider how the thermodynamic descriptions of the dierent phases in the system aect the dynamics of the simulation.
As was seen in this section, by solving the Cahn-Hilliard equation using the the thermodynamic models for the T i O system it was possible for the system to nd the
equilibrium compositions at each of the interfaces created. In principle, any function
for which common tangent constructions yield the same equilibrium compositions
should be equivalent and should yield the same interfacial arrangement as the one
shown here. One could, for example, linearly transform the Gibbs free energy function and still nd the same common tangents. The problem with this transformation
186
is that, unless the transformation involves just changing the reference state for the
elements in the system, the kinetics of the system will be aected since the diusion rates within each of the regions of the system are also dependent on the local
thermodynamics through:
ln O
DO 1 +
MO RT
=
ln XO
A linear transformation other than a constant shift of the Gibbs free energy curve
would yield a dierent thermodynamic factor, which multiplies the atomic mobility
to yield the chemical diusion coecient. In highly non-ideal systems, such as the
T i O system treated in this work, the thermodynamic correction is very important.
In summary, an accurate model for the system described here requires not only that
the Gibbs free energy functions used yield common tangent constructions compatible
with the equilibrium phase diagram but also accurate thermodynamic descriptions
that make it possible, provided diusion data is available, to relate the observed diffusion rates to atomic mobilities and to the compositional dependence of the chemical
potential of the diusion species.
A much more critical question is whether the model presented in this work is robust
enough with respect to small variations in the actual thermodynamic description of
the system. As can be seen in Fig. 7-6, the Gibbs free energy curves corresponding to
the bcc, hcp and f cc phases in the T i O system are very close to each other. A valid
question to ask is whether a small variation in the parameters used to describe the
thermodynamics of the system will have a signicant eect on the common tangent
construction that leads to the equilibrium phase diagram of the system. Moreover,
a change in the thermodynamic parameters used to describe the system may also
have a signicant eect on the heterogeneous Gibbs free energy regions, above the
equilibrium tie-lines and therefore lead to dierent values for the interfacial energies
of the system. By inspecting the results regarding the inuence of the value for the
gradient energy coecient ( see 7.6.1), k (see Eq. 7.13), on the dynamics of the
system, one may speculate that this last eect is not as important, although it is
necessary to perform a much more rigorous analysis. Although these questions are
beyond the scope of this thesis, it will be necessary to answer them in the near future
if this model is to be generalized to more complicated multi-component, multi-phase
systems.
7.10
7.10.1
Conclusions and Future Work

Conclusions for this Chapter
In this chapter, the Cahn-Hilliard equation was solved to describe the coupled growth
of multiple oxide phases during ceramic/metal joining operations. Although the
present paper focuses on the study of the thermochemical interactions between zirconiabased ceramics and pure titanium, it is possible to apply it to other similar systems.
187
With the methodology developed in this paper, it was possible to use real thermodynamic descriptions (based on the CALPHAD formalism) of the system and
incorporate them in a Phase-Field simulation in which the Cahn-Hilliard equation
is used to describe the kinetics of the phase transformations. By solving a single
kinetic equation it was possible to model the coupled growth of multiple phases without having to track the movement of every boundary in an individual manner. The
system was shown to satisfy the global mass conservation condition without actually
having to apply any local correction to the displacement of the individual boundaries.
It was also shown that an improper value for the gradient energy coecient could
lead to mass conservation problems which become more dominant as the interfacialarea/volume ratio of the system increases.
Through this chapter it was shown that the growth of the dierent phases is governed by bulk-diusion processes, following the square-root-of-time law. Furthermore,
it was demonstrated that the Stefan condition appears to be followed at all the moving
interfaces present in the system. From the one-dimensional simulations it was evident
that the actual rates of growth of the three dierent phases involved in the simulations were coupled through the thermodynamic properties at the moving interfaces.
Finally, it was demonstrated that the oxygen chemical is a monotonically-decreasing
function of distance, which is consistent with the fact that the Cahn-Hilliard equation
is based on the fundamental assumption of local thermodynamic equilibrium. From
the two-dimensional simulations, it was determined that, for these type of diusional
transformations, capillary eects due to local curvatures at the dierent moving interfaces do not play a role in determining the stable layering morphology to be expected
in these type of ceramic/metal interfacial reactions.
Although there is no experimental work reproducing the exact conditions that
were simulated in this chapter, it was still possible to establish the validity of the
assumptions made while developing the model, as well as the main ndings regarding
the qualitative behavior of the coupled layer growth simulation. The local equilibrium
hypothesis, which is central to the development of both sharp and diuse interface
models, has been determined to valid, at least for the temperatures of interest in this
work. Additionally, the fact that the dierent phase transformations occurring in the
T i O system are diusion-controlled has been veried by a series of experiments
on the oxidation of titanium and titanium alloys. Finally, the stability of planar
interfaces, with respect to undulated ones, seems to be conrmed by the available
literature.
Although there has been no attempt to make quantitative predictions regarding
the coupled growth of multiple oxide phases during ceramic/metal joining operations
in the T i O system, the model developed in this work seems to represent the relevant physics of the system in an acceptable manner. By relaxing the assumption of
constant mobility in this model, while using realistic values for the oxygen mobilities
in the dierent phases considered, it may be possible, in the future, to completely
188
validate the model and extend its applicability to other complex multi-phase moving
boundary problems of relevance not only in ceramic/metal interfacial interactions,
but in other material systems and technologies.
7.10.2
Future Improvements on the Model
Although the model presented in this chapter accounts for the formation and evolution of interfacial structures in this relatively simple system, there are a number of
improvements that need to be made in order for this model to the applicable to the
general problem of ceramic/metal interfacial formation and evolution. Two important improvements are listed below:
i) The model considers that the oxygen chemical potential on one side of the computational domain is xed at one particular value. Although this is a realistic
assumption in oxidation experiments, when zirconia-based ceramics interact with
titanium, the oxygen chemical potential at the zirconia/titanium interface is expected to change as the partial reduction progresses. This implies the incorporation in the present phase-eld model of a zirconia phase.
ii) Fig. 5-23 illustrates the interfaces observed when a zirconia substrate interacts
with a Cu Zr melt. In Section 5.7, it was shown that bulk thermodynamics
cannot explain the dissolution reaction observed. In that section, it was hypothesized that grain boundary attack by the Cu Zr melt could account for the
interface observed. If one were to reproduce this interface through phase-eld
modelling, it would be necessary to incorporate a model to account for the energy
related to the formation and evolution of grain boundaries. Such a model has
been successfully used by Bishop [165] in her study of Si3 N4 CaO inter-granular
structures. Incorporating the insights of that excellent work into the particular
problem of this interface could lead to a model for ceramic/metal dissolution
interactions.
189
190
Chapter 8
Conclusions and Future Work
The study of ceramic/metal interface formation and evolution is very important
from both an engineering and a scientic point of views. Better understanding of C/M
interface formation allows the improvement of important technological applications
that require the formation of synthetic C/M composites, such as solid oxide fuel cells,
ceramic-metal composites, ceramic-coated turbine blades and so forth. Additionally,
the diusional reactions and phase transformations occurring in these complex systems oer the opportunity to test several methodologies of analysis and develop new
models that can be used in the study of other equally complex phase transformation
systems.
8.1
General Conclusions
In this work, the study of zirconia/metal interfaces has been used as a model system to illustrate the use of basic thermodynamics and kinetics concepts to analyze
complex interfacial interactions. The selection of zirconia-based ceramics is not accidental, as it constitutes one of the ceramic families with numerous potential uses in
energy-related technologies, which will become increasingly important over time, due
to increased emphasis in eciency and clean energy access.
With regard to the formation of zirconia/active-metal interfaces, several conclusions can be made:
i) Perhaps the most general conclusion of this work is that it is possible to use
thermodynamic analysis to understand the complex diusional reactions and
phase precipitations occurring at ceramic/metal interfaces. Through the use of
metastable phase diagrams, activity maps and chemical equilibria it is possible to
understand the nature of zirconia/active-metal interfacial reactions under several
thermal and chemical conditions.
ii) In the particular case of zirconia/titanium reactions, it has been possible to establish, at least in a qualitative manner, the possible diusion paths that the system
has to take, based on simple assumptions regarding the kinetics of diusion and
191
dissolution.
iii) As has been shown in this work, it is the wide range of non-stoichiometry of
zirconia-based ceramics what mostly determines the diusion paths observed in
zirconia/titanium interfacial reactions.
iv) In the particular case of solid/solid zirconia/titanium interfacial interactions, the
relative size of the system plays an important role in the reaction sequences observed. When the amount of titanium in the system is comparable to that of
zirconia, the initial partial reduction of the zirconia lattice is followed by a complete decomposition of the ceramic when the maximum non-stoichiometry has
been reached. In cases where titanium lms are put in contact with zirconia substrates, this decomposition does not take place, and the diusion path observed
moves across the T i O binary system.
v) As mentioned in this thesis, the particular reaction product formed at the ceramic/braze interface determines the eectiveness of any given ceramic/metal
brazing procedure. By carefully designing the brazing alloy it is possible to precipitate wettable products that, put in contact with a liquid metal, promote
complete wetting of the ceramic substrate.
vi) Through the use of activity maps, it has been shown that the reaction products
observed in zirconia/active-braze interactions (the brazes could belong to the
Ag Cu T i or Cu T i Al systems) can be interpreted using the criterion of
monotonic decrease in the chemical potentials of the reacting species.
vii) Although there have been some attempts at analyzing the thermochemistry of
ceramic/metal joining prior to this work, the eect of the metal substrate eect on
the thermochemistry of the brazing alloy has not been quantitatively determined.
In this work, it has been shown that extensive solubility of the base metal in the
braze, coupled with high chemical anity between the base metal and the active
additive of the braze can aect, in a dramatic manner, the nature of the interfacial
interactions observed at the ceramic/braze interface.
viii) Besides the nature of the reaction products observed at ceramic/metal interfaces,
morphology plays an important role in determining the nal mechanical properties of ceramic/metal interfaces. By using simple thermodynamics and diusion
simulations it is not possible to gain insight into the nature of ceramic/metal
interfacial morphology and evolution. In this work, the Cahn-Hilliard equation
has been coupled to a CALPHAD model for the T i O system in order to study
the coupled oxide growth occurring at a zirconia/titaniun interface. Although
the model so far is only qualitative, several general results and implications of
this model have been validated by examining the available literature. Comparison to the analogous T i N system during nitriding processes suggests that the
model can be further used to understand other complex interfacial phenomena
important in several important technologies.
192
Within the limited scope of this work, a few contributions have been made that
not only can be applied to the analysis of the particular problem of this work, but
also to other important material-related problems. These contributions are mainly
through the development of thermodynamic models for several industrially important
systems:
i) A thermodynamic model of the Ag CuT i system has been developed through
the critical assessment of the existing experimental information. The fact that
the reaction products observed in zirconia/active-braze interactions in Section 5.5
can be understood using the results of this model provides additional support for
the model.
ii) A model for the Cu T i Zr system has also been assessed and the calculated
thermochemical properties of this system agree well with the experimental results. Although this model has not been explicitly used in this work, amorphous
alloys based on this system may be widely used as a high-temperature substitute
of Ag Cu T i systems.
iii) Despite having been studied and used for several decades, it was until this work
that a complete model for the Zr O system has been obtained. This model
is essential to understand the thermochemical reactions at ziconia/metal interfaces. Moreover, this model can be expanded to more complex systems that are
important as structural materials or solid oxide electrolytes for solid oxide fuel
cells, such as Y SZ, M GZ, CSZ, and so forth.
iv) The T i Zr O and Cu T i O systems have also been thermodynamically
assessed. Although the application of these models is somewhat constrained to
the study of zirconia/mtal interactions, the T i Zr O system may be used
in other studies that focus on the oxidation phenomena occurring in titanium
alloys.
8.2
Future Work
Even though it has been shown in this thesis that thermodynamic descriptions, based
on the CALPHAD methodology, can be used to understand the interfacial interactions observed in some zirconia/metal systems, there are several renements that
remain to be made and questions to be answered.
With regard to the thermodynamic models proposed in this work, several improvements remain to be made in some of the systems studied. For example, the
thermodynamic description of the T i Zr O system presents some discrepancies
with available experimental results, especially with regard to the experimental verication of several invariant reactions involving the liquid phase. Although these
discrepancies had no impact in the present work, they need to be resolved, especially
if this model is to be applied to other problems involving ionic melts with high titanium contents.
193
Considerable work is still required if one is interested not only in understanding

the nature of ceramic/metal interfacial interactions, but also in predicting and controlling them. Although this work indicates that it is feasible to predict the nature of
interfacial interactions for simple ceramic/metal systems, a more modest success has
been achieved in the prediction of such interactions in more complex systems, where
more components and phases play a dominant role.
The absence, in some zirconia/braze interaction experiments, of predicted phases,
such as the T i3 Cu3 O phase in zirconia/zirconia brazing operations may be due to
kinetic factors. However, this has not been proved in this work and further experimental and theoretical eort is needed if one is to resolve these discrepancies.
With regard to the evolution of ceramic/metal interfacial reaction products, a
need still exists for quantitative growth models, using information regarding the diffusion rates of the reacting species. Moreover, since many of these reaction products
are ionic in nature, the kinetic studies must incorporate coupled charge and mass
transfer mechanisms, which imply a more accurate model for the point defects in
these phases. The CALPHAD models used in this work do not, except for the model
for the T iO phase, incorporate charged species. Without this, a diusion model using
experimental data on ionic and electronic mobilities cannot be used.
An example of a technology where growth models for interfacial reaction products
is of fundamental importance is the use of zirconia-based ceramics as thermal barriers
for turbine blades. In these systems, high operating temperatures promote the interdiusion and reaction of oxygen and active elements within the alloy in the turbine
blade. The formation of an interfacial reaction product at the interface between the
blade and the thermal barrier eventually leads to the formation of interfacial cracks
and spalling of the protective zirconia barrier. In order to predict the operating life
of this system it is necessary to successfully account for the growth of the thermally
grown reaction products.
Although it has been shown that the Cahn-Hilliard equation can be successfully
coupled to a thermodynamic description of the likely interfacial reaction products
during zirconia/titanium interactions, the model is qualitative in nature, and quantitative validation is necessary. Real data on diusion rates are therefore necessary. In
principle, the model presented in this work my be applied to systems with a greater
number of components and eld variables other than composition, such as charge, for
example. Although this was not within the scope of this work, the incorporation of
these improvements may make this model applicable to other complex phase transformation and growth systems.
194
Appendix A
The Calphad Approach to
Thermodynamic Modelling:
Developing models for
Ceramic/Metal Systems
A.1
The CALPHAD Methodology: An Introduction
In order to facilitate the development of novel materials and processes it is essential to

understand the thermochemical behavior of the materials of interest as well as of the
equilibrium state that the system would reach under a particular set of conditions.
Because of the lack of comprehensive thermodynamic descriptions for many important material systems, the actual use of thermodynamics in materials design has been
limited to few materials families, such as steels, aluminum alloys, etc. Over the past
decades, and thanks to the pioneering work of Kaufman and Hillert and many more
researchers, a powerful methodology that enables the application of thermodynamics
to the development of material applications has slowly emerged. This methodology
is sometimes referred to as the CALPHAD Method and has as its main objective the
coupling of both thermochemistry and phase diagram information through the use of
semi-empirical models and computational techniques .
In general, it is possible to dene the Gibbs free energy of a phase using an
empirical function of the pressure, temperature and composition:
G = G f (P, T, Ci )
(A.1)
The particular choice of the model used to describe the thermodynamic properties of a phase depends on its chemical behavior and its internal structure . Over the
An excellent reference to this method can be found in the book by Saunders and Miodownik
Note that these empirical models do not contain information regarding the actual crystal struc-
[21]
195
years, models have been developed to describe random liquid and solid substitutional
and interstitial solutions, order/disorder transformations, phases with dened sublattice occupancies, intermetallic compounds, and so forth [21].
In this work, the general sublattice formalism is used to describe the thermodynamic properties of all the phases involved in the systems studied. In the remaining
sections of the present chapter the appropriate sublattice formalism notation will be
used to describe the underlying structure of the thermodynamic model for each of
the phases. Briey, a phase will be described using a similar expression to:
(A, B)N 1 (C, D)N 2
(A.2)
The model represented by Eq. A.2 indicates, for example, that the phase in question has two sub-structures (or sublattices) with dierent number of sites (N 1 and
N 2, respectively), and that the dierent components of such phase (A,B,C and D)
occupy distinct spatial regions within the phase.
The models developed using the formalism represented in Eq. A.2 are then thermodynamically optimized through a tting process whereby the adjustable coecients
in the Gibbs energy equations (such as Eq. A.1) are varied until the best (optimal)
representation of both thermochemical experiments and phase diagram determinations are obtained. By using procedures such as least-square minimizations [90] it
is possible to give this tting procedure a mathematical rigorousness. The optimizations and further thermochemical calculations in this work were performed with the
aid of the Thermo-Calc r program[144].
A.2
Calphad Models: A Brief Description
As noted in Section A.1, the development of models capable of accurately describing

the thermodynamic properties of chemical systems constitutes an essential aspect of
the CALPHAD methodology. In this section, a comprehensive description of the thermodynamic models used (and their mathematical formalism) in this work is presented
. In order to better illustrate the applicability of the models described, examples

using phases relevant to the systems of interest for this work will be presented.
When dealing with CALPHAD modelling, it is necessary to consider only two
main types of phases: stoichiometric compounds and solution phases (these solution
phases can then be classied as random, sublattice, ionic, etc.). The major dierence
ture of the phase. The term internal structure should be therefore understood as the arrangement of
the dierent species within the phase. Energetic relations between dierent structures are externally
imposed on the models, as adjustable coecients.
Measurements of chemical activities, enthalpies of formation, enthalpies of solution, etc.
Phase boundaries, invariant points, melting points, and so forth.
For a much more rigorous treatment on thermodynamic modelling and its relationship to the
CALPHAD methodology, one could refer to the work by Saunders and Miodownik [21].
196
between both types of phases is that a solution phase, contrary to a stoichiometric

compound, allows a nite solubility of more than one component or species within
the structural components of the phase. The remainder of this section contains a
brief description of the models used for the dierent kinds of phases assessed in this
work.
A.2.1
Stoichiometric Compounds
The integral Gibbs energy of a pure species or stoichiometric compound G (in this
ij
case a binary ij compound) is given by the equation:
G H SER = 0 Hij,f T 0 Sij,f + GHSERi + GHSERj

ij
(A.3)
where 0 Hij,f is the standard enthalpy of formation for the compound ij (that has
the crystal structure ), 0 Sij,f is the entropy of formation, and GHSER can in turn
be dened as Gi HiSER , where HiSER is the enthalpy of the element i in its dened
reference state at 298.15 K.
A.2.2
Random Substitutional Phases
Random substitutional models are used with phases whose components can mix on
any spatial position available to that phase (in a crystal structure, it is possible to
think of lattice sites, for example). In general, the Gibbs energy G of such a phase
can be represented by
G = 0 G + id G + xs G
m
m
(A.4)
where 0 G represents the contribution of the pure components of the phase to the total Gibbs energy, id G is the ideal mixing contribution (considering non-interaction
m
components) and xs G represents any interaction between components within the
m
phase.
Since there is no interaction energy between components in an ideal solution, the
ideal mixing energy is only determined by the congurational entropy of the solution:
id
x ln x
i
i
G = RT
m
i
where xi is the molar fraction of component i in phase .
substitutional solid solutions, simple liquids, gas phase
197
(A.5)
In an ideal solution, xs G = 0. In most cases, however, it is necessary to consider

m
the energetic interactions between components within the phase. One of the rst
attempts to use a mathematical description for the excess Gibbs energy of a phase
was that of Van Laar in 1908 [166]:
G,xs = xi xj Lij
mix
(A.6)
where Lij is the interaction parameter between components i and j of the solution.
In general Lij can be composition- (higher order terms are necessary) as well as
temperature- (there is an excess entropy of mixing) dependent. In the case where there
are more than two components and the higher order interactions become important,
additional interaction parameters can be added to the expression for the Gibbs energy.
The general expression for a random solution (in this case, a third order solution) is
then given by:
n1
x 0 G +RT
i
i
G =
i
x
i
ln
x
i
x x L + x x
i j i,j
i j=i=k xk=j=i Li,j,k (A.7)
j=i+1
where L can be also a function of temperature and composition:

ij
L =
ij
a + kb T
x x
i
j
(A.8)
In cases when the Gibbs free energy of a ternary system is highly asymmetrical ,
it may be necessary to have a ternary interaction parameter with greater degrees of
freedom if an accurate description of the energy of the ternary system is desired. In
order to do this, the ternary term appearing in Eq. A.7 would be given by:
3
G = xi xj xk vi L + vj L + vk L
i,jk
j,ik
k,ij
(A.9)
where vi is given by:
1 xi xj xk
(A.10)
3
Note that vi is identical to xi when the system is a ternary. In the case of a
higher-order system that is not the case. Additionally, the sum over all vi is unity in
all cases, while the sum of xi is only unity in the case of a ternary system.
vi = xi +
Perhaps the best example of a random substitutional phase is a metallic liquid solution. An important random solution phase assessed in this work, for example, is the
liquid Ag Cu T i phase. This phase has very interesting properties (the existence
of a ternary miscibility gap) that dene clear operational limits for ceramic/metal
Take for example a case were two of the binary subsystems have strong negative deviations
from ideal behavior while the third sub-system has strong positive deviations.
198
joining techniques using such alloys.
A.2.3
The Sublattice Model
Although the random solution model described in A.2.2 is very useful and widely
applicable, it has some limitations, for example:
This model cannot take into account preferential site occupation in intermetallic
phases .
The random solution model does not constitute a realistic representation of
solid interstitial solutions since interstitial atoms occupy clearly distinct sites in
the solution .
The complex behavior of ionic melts, where charge neutrality must be taken
into account as a further constraint to the system, cannot be described using
random solution models .
The limitations listed above come from the fact that, in the random solution
formalism, all regions of space must be treated equivalently. It is obvious that to
overcome its limitations, it is necessary to expand this model so unlike types of sites
within a phase are treated dierently . Such a model was developed some 20 years
ago, by Sundman and gren [167]. This model, appropriately called the sublattice
A
model, relies on the explicit dierentiation of the structural components within a
phase (sublattices) . Thanks to its practical success, this formalism has become one
of the principal methods used to describe a wide range of phases with diverse chemical
and physical behaviors [21]. Its exibility allows its application to the description of
phases with widely diering behavior, from solid interstitial solution phases to ionic
liquid solutions, to phases undergoing order/disorder transformations, and so forth.
Fig. A-1 shows an example of a crystalline phase having two distinct sublattices.
In this particular example, we have four species (A,B,C,D), with species A and B
allowed to occupy only the body-centered sites while species C and D occupy the corner sites. Note that although in the gure we can distinguish between species in each
sublattice, a central characteristic of the sublattice model is that the species within
See for example the Wagner-Shottky model used in Hari Kumar et al. [30] to describe the CuT i
and Cu4 T i phases in the Cu T i system.
For example the phase in the ZrO system [52]. In this phase, oxygen occupies the interstitial
sites of the BCC structure.
The two-sublattice ionic liquid model has been used to describe the liquid phase in the Zr O
system assessed by Arroyave et al. [52].
The distinction among dierent sites can be articial, like in an ionic melt or natural, like in
an interstitial solution.
It is important to note, however, that this model is phenomenological in nature and it does not
attempt to represent structural relationships between the dierent sublattices present in any given
phase (crystalline or non-crystalline).
These species could be elements, stable molecules, metastable complexes, etc.
199
each sublattice behave as if they were in a random solution, that is, they randomly
mix. Using the sublattice model, it is possible to describe chemical interactions between species located in dierent sublattices (through the Compound Energy Method
[168]) as well as within the same sublattice. Therefore it is possible to represent the
behavior of complex phases.
In the remainder of this subsection, a brief description of the mathematical formalism of the general Sublattice Model is presented. The original description of this
model was published by Sundman and gren [167]. A general overview of the appliA
cation of the Sublattice Model to the description of several types of solutions is given
in Saunders and Miodownik [21].
Figure A-1: Body-centered crystalline structure. Observe that there is preferential occupation of atoms in two distinct sites: blue (A) and black (B ) species in body-centered sites,
red (C ) and green (D) species in corner sites.
A.2.3.1
Mathematical Description of Sublattice Model
Perhaps one of the most important things to understand about the sublattice formalism is its notation. We can take, as an example, the phase depicted in Fig. A-1. In
the sublattice formalism, such a phase is described as
(A, B)N 1 (C, D)N 2
(A.11)
Eq. A.11 can be understood as follows: the phase has two distinct sublattices (or
clearly dened crystallographic positions). The number of sites in sublattices 1 and
2 are N 1 and N 2 , respectively. The rst sublattice is occupied by species A and B.
The second sublattice is occupied by species C and D.
200
The occupancy of each sublattice can be dened using the site-fraction matrix Y
dened by Sundman and gren [167]:
A
Y =
1
1
1
1
yA yB yC yD
(A.12)
2
2
2
2
yA yB yC yD
s
where yi corresponds to the site fraction of species i (ns is the mole number of i in subi
s
s
lattice s) in sublattice s (N s is the total number of moles in sublattice s) (yi = ni/N s
1
1
2
2
). Note that for this particular case yC = yD = yA = yB = 0.
Following a description analogous to that of a random solution (see Eq. A.4 ), the
Gibbs energy of a sublattice phase is given by:
G = ref G + id G + xs G
m
m
m
(A.13)
The rst term ref G corresponds to the Gibbs energy reference state. This refm
erence state corresponds to the state at which there is complete site occupation by
only one of the species in each sublattice:
ref
1 2
G = yA yC 0 G
(A)
1 (C)1
1 2
+yA yD 0 G
(A)
1 (D)1
1 2
+yB yC 0 G
(B)
1 (C)1
1 2
+yB yD 0 G
(B)
1 (D)1
(A.14)
where 0 G (j) corresponds to the Gibbs free energy of the reference compound ij in
(i)1 1
the crystal structure analogous to the expression found in Eq. A.3. The Gibbs free
energy of the reference compound represents a chemical interaction between atoms
occupying dierent sites within the structure. Eq. A.14 can be generalized using:
ref
PI0 (Y ) 0 G
I0
G =
(A.15)
where PI0 (Y ) corresponds to the zeroth-order product of site fractions (see Eq. A.12)
corresponding to each of the reference compounds [167].
As noted above, a central aspect of the sublattice model is the fact that the species
within each sublattice behave as if they were part of a random solution. Therefore,
the ideal contribution to the total Gibbs free energy of a sublattice phase is given by
the congurational entropy contribution of each of the sublattices (see Eq. A.5):
id
G = RT
m
Ns
s=1,2
s
s
yi ln (yi )
i=A,B,C,D
201
(A.16)
The excess Gibbs energy of a sublattice phase is given by the chemical interactions between species within each of the sublattices. In general, we can think that
chemical interactions among species within the same sublattice arise by the fact that
such species have nearest neighbors in the sublattice they are in[169]. Without any
interaction energy, the solutions within each sublattice are ideal.
For the phase depicted in Fig. A-1, we can assume for example that there is a
non-ideal chemical interaction between species C and D in the second sublattice when
A or B are present in the rst sublattice. For this situation, the excess Gibbs energy
is given by:
ex
1 2 2
1 2 2
G = yA yC yD L
m
A:C,D + yB yC yD LB:C,D
(A.17)
where L
A:C,D can be a function of both temperature and composition, as in Eq. A.8.
This case can be generalized using the equation:
ex
Gm =
PIZ (Y ) LIZ
(A.18)
Z>0 IZ
where Z is the order of the array necessary to dene the interactions of arrangement
I. For example, the interaction A : C can be dened by a zeroth-order array and
1 2
denes the Gibbs energy of compound (A)1 (C)1 . PI0 will be given then by yA yC .
When the interaction arrangement I is A : C, D, then we are dening a chemical
interaction between species C and D in the second sublattice, while A is present in
1 2 2
the rst one (L
A:C,D ). In this case PIZ would be P(A):(C,D)1 or yA yC yD .
Using Eq. A.14, Eq. A.16 and Eq. A.14, we have an expression for the total Gibbs
energy for the phase (A, B)1 (C, D)1 :
1 2
1 2
1 2
1 2
G = yA yC 0 G + yA yD 0 G + yB yC 0 G + yB yD 0 G
B:D
A:C
A:D
B:C
Ns
+ RT
s=1,2
s
s
yi ln (yi )
(A.19)
i=A,B,C,D
1 2 2
1 2 2
+ yA yC yD L
A:C,D + yB yC yD LB:C,D
We can generalize the expression for any sublattice phase, with any number of
components and sublattices as [167]:
PI0 (Y ) 0 G + RT
I0
G =
I0
s
s
yi ln (yi ) +
Ns
s
PIZ (Y ) LIZ
(A.20)
Z>0 IZ
In order to make equilibrium calculations it is important to be able to derive

mathematical expressions for partial quantities (i.e. chemical potentials). Using the
202
methodology described in Hillert [168, pp. 70-72], it is possible to obtain, from

the extensive quantity A(T, P, Ni ) the partial quantity Aj (A/Nj )T,P,Nk . Using
A = N Am , we have:
Aj = Am +
Am
xj
xi
Am
xi
(A.21)
If one were to obtain a mathematical expression for the chemical potential of A

in the phase (A) (A, C), given the fact that the total Gibbs energy expression for this
phase would be:
2
2
2
2
2
2
G = yA 0 GA:A + yB 0 GA:B + RT [yA ln (yA ) + yB ln (yB )]
2 2
+yA yB 0 LA:A,B
(A.22)
For this particular case, the chemical potential of A, A , is equivalent to the partial
quantity GA:A . Using Eq. A.21, the following expression is obtained:
2
2
2
A = GA:A = 0 GA:A + RT yA ln yA + yB 0 LA:A,B
A.2.4
(A.23)
Applications of the Sublattice Model
As noted above, the sublattice model is very powerful and versatile, as it can be
applied to a broad range of phases types. A good example of the application of the
sublattice model to intermetallic phases is the description of the M T i2 phases in the
assessment of the Cu T i system by Hari Kumar et al. [30]. In this system the
phases CuT i2 and AgT i2 (M oSi2 is their prototype structure) form a solid solution
with each other and therefore it is reasonable to assume that they can both be represented by the same phase:
(Ag, Cu)1 (T i)2
(A.24)
According to Eq. A.24, Ag and Cu occupy the rst sublattice of the structure
M oSi2 . Both elements would form a random solution in the rst sublattice, with T i
always present in the second sublattice. The Gibbs energy of this phase would be
given by:
1
1
GM T i2 = yAg 0 GM T i2 + yCu 0 GM T i2i
Ag:T i
Cu:T
1
1
1
1
+ RT yAg ln yAg + yCu ln yCu
(A.25)
1
1
+ yAg yCu LM T i2 i
Ag,Cu:T
Note that, in the sublattice formalism, it is not possible to directly dene a chemical potential
for a pure element i. It should be derived (when possible) from the partial Gibbs energy of the
components in the phase.
203
where 0 GM T i2 corresponds to the Gibbs energy of the end-member AgT i2 and 0 GM T i2i
Ag:T i
Cu:T
is the Gibbs energy of the pure CuT i2 compound. This expression can be further
1
1
simplied using the constraint yCu + yAg = 1
Another example of an application of the sublattice model is the description of
interstitial solid solutions. In these phases, small elements such as O, C, N , H occupy
the interstitial sites of the structure. The system Zr O presents two such phases:
(bcc) and (hcp). In these phases, oxygen occupies the interstitial sites and their
description is given by:
(Zr)1 (O, V a)c
(A.26)
where c corresponds to the ratio between interstitial and substitutional sites in the
structure. In this system, c has a value of 3 and 0.5 for the bcc and hcp phases,
respectively. In order to properly describe this phase, it is necessary consider the
presence of vacancies in the interstitial sites. The Gibbs energy of this phase can be
represented using
2
2
G, = yV a 0 G, a + yO 0 G,
Zr:O
Zr:V
2
2
2
2
+ c RT yO ln yO + yV a ln yV a
2 2
+ yO yV a L, a
Zr:O,V
(A.27)
The reference compounds for this phase would then be 0 G, a and 0 G, . 0 G, a

Zr:V
Zr:O
Zr:V
corresponds to the Gibbs energy of the pure bcc or hcp phases. 0 G, in turn corZr:O
responds to the Gibbs energy of the hypothetical compounds ZrO3 or ZrO0.5 which
correspond to a state at which there is full occupation of oxygen atoms in the interstitial sites . In this model, the L interaction parameters describe the chemical
interaction between species within the same sublattice . The G Gibbs free energy
expressions correspond to the chemical interactions between species belonging to different sublattices.
The power of the sublattice model can be best illustrated by considering what
would happen if a third element, say T i, was introduced into the system. In this
case, the sublattice representation of the phase would be
(T i, Zr)1 (O, V a)c
(A.28)
Providing there was already an expression for the total Gibbs energy of the
(T i)1 (O, V a)c and (T i, Zr)1 (V a)c systems, the total Gibbs energy for the ternary
system described in Eq. A.28 (see Section 4.5) would be given by:
In interstitial solutions of group IV elements with small gaseous species such as oxygen, hydrogen
and nitrogen, the gas atoms occupy the octahedral interstitial sites [58].
Such full occupation is observed in the case of the hcp solid solution.
In this case, vacancies and oxygen interstitials.
204
1 2
2
1 2
2
1
1
G, = yZr yV a 0 G, a + yZr yO 0 G, + yT i yV a 0 G, a + yT i yO 0 G,
T i:O
T i:V
Zr:O
Zr:V
+ RT
1
1
1
1
yT i ln yT i + yZr ln yZr
2
2
2
2
+ c yO ln yO + yV a ln yV a
1
2 2
1 2 2
+ yZr yO yV a L, a + yT i yO yV a L, a
Zr:O,V
T i:O,V
(A.29)
1
1 2
1 1 2
+ yZr yT i yO L, i:O + yT i yT i yV a L, i:V a
Zr,T
Zr,T
Finally, it is possible to use the sublattice formalism to describe the thermodynamic behavior of ionic melts [170]. This model considers two types of sublattices.
The electropositive ions are contained in one of the sublattices, while electronegative
ions and neutral atoms are located in the other sublattice. This model also considers
the hypothetical presence of charged vacancies so the composition of the melt can
deviate towards the pure melt of the element with electropositive character, maintaining, at all times, the neutrality condition of the liquid. The sublattice formula
for this model is represented by:
v
Ci+vi
0
Aj j , V a, Bk
(A.30)
where C are the cations, A anions, V a (hypothetical) vacancies and B neutrals.

The charges of the ions are represented by vi and vj . In order to maintain electroneutrality throughout the entire compositional range of the melt, it is necessary to
vary the values for P and Q:
P =
j
Q=
i
vj yAj + QyV a
(A.31)
vi yCi
From Eq. A.31 it is obvious that the hypothetical vacancies will have an induced
(negative) charge of - |Q|. If neutral atoms in the anion sublattice are excluded and
only one type of anion in the melt (for example oxygen) the Gibbs energy of this
phase would be given by:
G=
i
yCi yAi 0 GCi :Aj + QyV a
+RT P
i
+
i1
i2
yi1 yi2 yj Li1 ,i2 :j +
i1
i2
yAj ln yAj + yV a ln (yV a )

(A.32)
2
yi1 yi2 yV a Li1 ,i2 :V a
yCi ln (yCi ) + Q
yCi 0 GCi
yi yj yV a Li:j,V a
where 0 GCi :Aj is the Gibbs energy of formation for the corresponding compound and
205
GCi is the Gibbs energy of the pure liquid cation. The L interaction parameters, can
be dened as:
Li1 ,i2 :j represents the interaction between cations i1 and i2 in the presence of
a common anion; for example LZr+4 ,T i+4 :O2 represents the interaction term in
the (liquid) system ZrO2 T iO2 (see Section 4.5).
Li1 ,i2 :V a represents interactions between metallic elements; for example LZr,T i:V a
in the T i Zr binary system.
Li1 :j,V a represents interactions between an anion and the hypothetical vacancies
in the presence of a cation i1 ; for example LZr+4 :O2 ,V a (see Section 4.2).
An example of a phase assessed in this work would be the ionic liquid in the ZrO
system (Section 4.2):
Zr+4
2
42yO
O2 , V a
(A.33)
The Gibbs energy of this phase is given by:
Gionic = yZr+4 yO2 0 G(Zr+4 )2 :(O2 )4 + 4yZr+4 yV a4 0 G(Zr+4 ):(V a4 )

+RT [(4 2yO2 ) yZr+4 ln (yZr+4 )]
+RT [4 {yO2 ln (yO2 ) + yV a4 ln (yV a4 )}]
+yZr+4 yO2 yV a4 LZr+4 :O2 ,V a4
206
(A.34)
Appendix B
Derivation of an expression for the
Chemical Potential of an element i
in a Sublattice Phase
In principle, the partial Gibbs energy with respect to a component i in a phase
is equivalent to its chemical potential and is dened as:
Gi = (i)
G
ni
(B.1)
T,P.nj
In a phase that has sublattices, however, it is impossible to vary the content of

one component alone, as Eq. B.1 requires, unless it is present in all the sublattices.
This diculty present in all phases with more than one sublattices can be overcome
if one, instead of dening the chemical potential of an element, denes the chemical
potential of a compound:
Gi j = (ij)
G
ni j
(B.2)
T,P.nj k
According to [167], it is possible to dene the chemical potential of component I0:
I0 = G +
s
s
yi
s
yj
j
G
s
yj
(B.3)
The second term in Eq. B.3 corresponds to the sum of the partial derivatives of the
expression of the Gibbs energy of the phase with respect to the site fractions of the
elements present in each of the compounds sublattice. The third term corresponds
to all the partial derivatives of the Gibbs free energy with respect to the site fraction
of all the components of the phase in all the sublattices.
207
As a more concrete example, it is possible to use the model (A)a (B, C)b to calculate the chemical potential of compound (A)a (B)b :
Aa Bb = G +
G G
G
G
G
+
yA
yB
yC
yA
yB
yA
yB
yC
(B.4)
Due to the stoichiometry constrain, it is not possible to dene a chemical potential for element A, as it is not present in all the sublattices of phase . However, the
chemical potential of the compounds in phase and the chemical potential of the
elements in phase are related through:
Aa Bb = aA + bB
(B.5)
Aa Cb = aA + bC
According to Eq. B.5, the chemical potential of B can be obtained using a linear
combination of the chemical potentials of the components Aa Bb and Aa Cb :
Aa Bb Aa Cb = aA + bB aA + bC
B =
1
b
(B.6)
(Aa Bb Aa Cb bC )
Thus, the chemical potential of B can be dened with respect to that of C. C is

any constituent of the sublattice. If the sublattice in question has vacancies, and it
is assumed that such point defects are at thermal equilibrium, it can be considered
that their chemical potential is zero and the expression for the chemical potential is
greatly simplied.
208
Appendix C
Model Parameters
In this appendix, the model parameters for the phases that were thermodynamically assessed in Chapter 3 and Chapter 4 will be presented. The expressions for the
Gibbs free energy of formation of the dierent phases of the pure components will not
be included, as they are available in the SGTE thermodynamic databases. Unless
otherwise stated, the Gibbs free energy of formation of stoichiometric compounds is
also taken from the SGTE thermodynamic databases. Parameters marked with
were obtained in this work. For a detailed description of the CALPHAD models used
in the optimization, refer to Chapter A. Details of the assessments are available in
Chapter 3 and Chapter 4.
Table C.1: Thermodynamic parameters for the ionic liquid phase.

(Ag +1 , Cu+1 , Cu+2 , T i+2 , T i+3 , T i+4 )P (O2 , V a)Q
Ag Cu [29]
0 liq
LAg+1 ,Cu+1 :V a
+17, 534.6 4.45479T
Lliq +1 ,Cu+1 :V a
Ag
+2, 251.3 2.6733T
Lliq +1 ,Cu+1 :V a
Ag
+492.7
Ag T i
Lliq +1 ,T i+2 :V a
Ag
+25, 632
Lliq +1 ,T i+2 :V a
Ag
327
Cu T i [30]
Lliq +1 ,T i+2 :V a
Cu
19, 330 + 7.651T
Lliq +1 ,T i+2 :V a
Cu
0
continued on next page
209
continued from previous page

2
Lliq +1 ,T i+2 :V a
Cu
+9, 382 5.448T

Ag Cu T i
0 liq
LAg+1 ,Cu+1 ,T i+2 :V a
+32, 400
Cu Zr [48]
Lliq +1 ,Zr+4 :V a
Cu
61, 685.53 + 11.2924T
Lliq +1 ,Zr+4 :V a
Cu
8, 830.66 + 5.04565T
Cu T i Zr [26]
0 liq
LCu+1 ,T i+2 ,Zr+4 :V a
+23, 828.5
Lliq +1 ,T i+2 ,Zr+4 :V a

Cu
28, 081.4
Lliq +1 ,T i+2 ,Zr+4 :V a

Cu
+23, 828.5
T i Zr [49]
Lliq+2 ,Zr+4 :V a
Ti
968
Cu O [71]
Gliq +2 :V a
Cu
Gliq + 600, 000

Cu
Gliq +1 :O2
Cu
3 GCU 2OLIQ
Gliq +2 :O2
Cu
4 GCU 2OLIQ 44, 058 + 25T
0 liq
LCu+1 :O2 ,V a
+27, 004 + 2.6T
Lliq +1 :O2 ,V a
Cu
9, 894 + 5.73T
Lliq +1 :O2 ,V a
Cu
20, 462 9.8T
Lliq +1 ,Cu+2 :O2

Cu
6, 879
Lliq +1 ,Cu+2 :O2

Cu
8, 000
T i O [34]
Gliq+3 :V a
Ti
Gliq + 2 104
Ti
Gliq+4 :V a
Ti
Gliq + 4 104
Ti
Gliq+2 :O2
Ti
2 GT IOA + 41, 2507 201.62T
Gliq+3 :O2
Ti
GT I2O3 + 190, 919 71.49T
Gliq+4 :O2
Ti
2 GT IO2 + 178, 003 62.4769T

210

0
Lliq+2 :O2 ,V a
Ti
249, 324 + 112.42T

Zr O [52]
0 liq
GZr+4 :O2
0 liq
LZr+4 :O2 ,V a
2 GZRO2A + 198, 026 71.62T
Lliq +4 :O2 ,V a
Zr
+40, 050
+136, 748 67.01T
T i Zr O
0
Lliq+4 ,Zr+4 :O2

Ti
105, 027
Lliq+4 ,Zr+4 :O2

Ti
+39, 985
Lliq+4 ,Zr+4 :O2 ,V a

Ti
30, 010
Lliq+4 ,Zr+4 :O2 ,V a

Ti
+49, 950
Table C.2: Thermodynamic parameters for the bcc phase.

(Ag, Cu, T i, Zr) (O, V a)3
Ag Cu [29]
0
Lbcc
Ag,Cu:V a
35, 000 8T
Ag T i
Lbcc i:V a
Ag,T
22, 900 + 0.5T
Lbcc i:V a
Ag,T
1, 500
Cu T i [30]
Lbcc i:V a
Cu,T
3, 389
Cu Zr [48]
Lbcc
Cu,Zr:V a
7, 381.13
T i Zr [49]
Lbcc
T i,Zr:V a
4, 346 + 5.489T
Cu T i Zr [26]
Lbcc i,Zr:V a
Cu,T
12, 000
211
(Ag, Cu, T i, Zr) (O, V a)3

Cu O
0
Gbcc
Cu:O
+1 106 + 3 GHSER + GHSER

O
Cu
T i O [34]
0
0
Gbcc
T i:O
4 105 + 3 GHSER + GHSER

O
Ti
Lbcc a
T i:O,V
1, 207, 294 + 274.32T

Zr O [52]
Gbcc
Zr:O
405, 010 + 3 GHSER + GHSER

O
Zr
Lbcc a
Zr:O,V
1, 237, 630 + 278.01T
0
0
T i Zr O
0
Lbcc
T i,Zr:O
866, 900 + 329T
Lbcc
T i,Zr:O
250, 000 + 146T
Table C.3:
phase.
Thermodynamic parameters for the f cc
(Ag, Cu, T i, Zr) (O, V a)

Ag Cu [29]
0 f cc
LAg,Cu:V a
+33, 819 8.124T
Lf cc
Ag,Cu:V a
5, 601.9 + 1.32997T
Ag T i
0
Lf cc i:V a
Ag,T
23, 405
Cu T i [30]
Lf cc i:V a
Cu,T
9, 882
Lf cc i:V a
Cu,T
15, 777
Cu Zr [48]
Lf cc
Cu,Zr:V a
+2, 233
T i Zr [49]
212
(Ag, Cu, T i, Zr) (O, V a)

0
Lf cc
T i,Zr:V a
4, 346 + 5.489T
Cu T i Zr [26]
Lf cc i,Zr:V a
Cu,T
0
Cu O
Gf cc
Cu:O
+1 106 + GHSER + GHSER

O
Cu
Lf cc a
Cu:O,V
1, 017, 730 + 29.6T
0
0
T i O [34]
0 f cc
GT i:O
0 f cc
LT i:O,V a
+59741 + GT IOA
11, 628 + 4.99T
Zr O [52]
Gf cc
Zr:O
N/A
0 f cc
LZr:O,V a
N/A
Table C.4:
phase.
Thermodynamic parameters for the hcp
(Ag, Cu, T i, Zr) (O, V a)0.5

Ag Cu [29]
0
Lhcp
Ag,Cu:V a
35, 000 8T
Ag T i
0 hcp
LAg,T i:V a
50, 050
Lhcp i:V a
Ag,T
27, 000
Cu T i [30]
0
Lhcp i:V a
Cu,T
16, 334
Cu Zr [48]
Lhcp
Cu,Zr:V a
+11, 337
T i Zr [49]
213
(Ag, Cu, T i, Zr) (O, V a)0.5

0
Lhcp
T i,Zr:V a
+5, 133
Cu T i Zr [26]
Lhcp i,Zr:V a
Cu,T
0
Cu O
Ghcp
Cu:O
+1 106 + GHSER + 0.5 GHSER

O
Cu
T i O [34]
Ghcp
T i:O
260, 898 + 33.86T + GHSER + 0.5 GHSER

Ti
O
Lhcp a
T i:O,V
11, 628 + 4.99T
0
0
Zr O [52]
Ghcp
Zr:O
286, 672 + 43.99T + GHSER + 0.5 GHSER

Zr
O
Lhcp a
Zr:O,V
40, 056 + 19.01T
Lhcp a
Zr:O,V
9, 995
T i Zr O
0
Lhcp
T i,Zr:O
+20, 000
Lhcp
T i,Zr:O
118, 963 77.7T
0 hcp
LT i,Zr:O,V a
25, 000
Table C.5: Thermodynamic parameters for the CuT i

phase. Cu T i parameters taken from [30]
(Ag, Cu, T i) (Ag, Cu, T i), = Ag, Cu, T i
0
GCuT i
Cu:Cu
+10, 000 + 2 GHSER

Cu
GCuT i
T i:Cu
+42, 412 6.544T + GHSER + GHSER

Cu
Ti
GCuT ii
Cu:T
22, 412 + 6.544T + GHSER + GHSER

Cu
Ti
0
0
i
GCuT i
T i:T
GCuT i
Ag:Ag
GCuT i
T i:Ag
+10, 000 + 2 GHSER

Ti
+10, 000 + 2 GHSER
Ag
+24, 080 1.32T + GHSER + GHSER
Ag
Ti
214
(Ag, Cu, T i) (Ag, Cu, T i), = Ag, Cu, T i

0
GCuT ii
Ag:T
4, 080 + 1.32T + GHSER + GHSER

Ag
Ti
GCuT i
Cu:Ag
25, 000 + GHSER + GHSER

Ag
Cu
GCuT i
Ag:Cu
25, 000 + GHSER + GHSER

Ag
Cu
LCuT ii:
Cu,T
15, 419
LCuT i i
:Cu,T
15, 578
LCuT ii:
Ag,T
23, 000
LCuT i i
:Ag,T
23, 000
LCuT i
Cu,Ag:Ag
50, 000
LCuT i
Cu,Ag:Cu
50, 000
0
1
LCuT i i
Cu,Ag:T
21, 000
LCuT i
Cu,Ag:Cu
3, 500
LCuT i
:Cu,Ag
50, 000
Table C.6: Thermodynamic parameters for the Cu4 T i

phase. Cu T i parameters taken from [30]
(Cu, T i)4 (Cu, T i)
0
+25, 000 + 5 GHSER

Cu
Cu4 T
GT i:Cui
80, 000 11.693T + GHSER + 4 GHSER

Cu
Ti
Cu4 T
GCu:T ii
30, 055 + 11.693T + 4 GHSER + GHSER

Cu
Ti
0
0
Cu4 T i
GCu:Cu
Cu4
GT i:TT i
i
+25, 000 + 5 GHSER

Ti
LCuT ii:Cu
Cu,T
17, 089
0
0
LCuT ii:T i
Cu,T
17, 089
LCuT i i
Cu:Cu,T
15, 767
LCuT i i
T i:Cu,T
15, 767
215
Table C.7: Thermodynamic parameters for the CuM2

phase. Cu T i parameters taken from [30]. Cu Zr
parameters taken from [48]
(Ag, Cu) (T i, Zr)
0
GCuM2
Cu:T i
36, 393 + 14.064T + GHSER + 2 GHSER

Cu
Ti
GCuMi2
Ag:T
8, 325 + 3.3T + GHSER + 2 GHSER

Ag
Ti
GCuM2
Cu:Zr
43, 904 + 5.1905T + GHSER + 2 GHSER

Cu
Zr
GCuM2
Ag:Zr
15, 000 + 2.6T + GHSER + 2 GHSER

Ag
Zr
LCuM2 i
Ag,Cu:T
14, 500
LCuM2
Cu:T i,Zr
+30, 250
LCuM2
Cu:T i,Zr
6, 205
LCuM2
Cu:T i,Zr
+1, 028
Table C.8: Thermodynamic parameters for stoichiometric intermetallic compounds.

Stoichiometric intermetallics
Cu T i [30]
GCu2 Tii
Cu:T
17, 628 + 2 GHSER + GHSER

Cu
Ti
Cu3 T
GCu:T ii2
46245 + 10.86T + 3 GHSER + 2 GHSER

Cu
Ti
Cu4 T
GCu:T ii3
68, 236 + 15.946T + 4 GHSER + 3 GHSER

Cu
Ti
Cu Zr [48]
0
GCuZr
Cu:Zr
2.01 10+4 7.63196T + GHSER + GHSER

Cu
Zr
GCu5 Zr
Cu:Zr
6.1794 10+4 + 5 GHSER + GHSER

Cu
Zr
GCu8 Zr3
Cu:Zr
1.48063 10+5 + 8 GHSER + 3 GHSER

Cu
Zr
GCu51 Zr14
Cu:Zr
8.43412 10+5 + 51 GHSER + 14 GHSER

Cu
Zr
Cu10 Zr
GCu:Zr 7
2.4174 10+5 + 10 GHSER + 7 GHSER

Cu
Zr
0
0
0
Cu T i Zr [26]
0
iZr
GCu2 Ti:Zr
Cu:T
7.6377 104 + 2.0206882 101 T + 2 GHSER + GHSER + GHSER

Cu
Ti
Zr
216
Table C.9: Thermodynamic parameters for the T iO

phase. [34]
(T i+2 , T i+3 , V a) (T i0 , V a) (O2 )
0
iO
GT+2 :T i:O2
Ti
259, 159 52.3756T + 2 GHSER + GHSER

Ti
O
iO
GT+3 :T i:O2
Ti
247, 865 + 2 GHSER + GHSER

Ti
O
GT iOi:O2
V a:T
GT IOA + 237, 249 3.0751T
iO
GT+2 :V a:O2
Ti
GT IOA + 59, 741 3.30751T
0 T iO
GT i+3 :V a:O2
0 T iO
GV a:V a:O2
0 T iO
LT i+2 :T i,V a:O2
0.5 GT I2O3 + 45, 030 + 6.2254T
iO
LT+2 ,T i+3 :V a:O2
Ti
173, 492
0
0
0
472, 450 + 185.7944T
Table C.10: Thermodynamic parameters for the T iO2

(rutile) phase. T i O parameters from [34]
(T i+4 , Zr+4 ) (O2 , V a2 )
0
0
GT iO2:O2
T i+4
+GT IO2 + 22, 057.2
GT iO2 :O2
Zr+4
+GZRO2A + 30, 971 4.02T
0 T iO2
GT i+4 :V a2
0 T iO2
GZr+4 :V a2
0 T iO2
LT i+4 ,Zr+4 :O2
+GZr
HSER + 40, 000
LT iO2:O2 ,V a2
T i+4
150, 570.4
+GT i
HSER + 40, 000
18, 625 + 4T
Table C.11: Thermodynamic parameters for the ZrO2

phase [52].
(Zr+4 ) (O2 , V a2 )2
0
2
GZrO:O2
Zr +4
+GZRO2A
217
(Zr+4 ) (O2 , V a2 )2
0
0
2
GZrO:V a2
Zr+4
+GZr
HSER + 200, 000
LZrO2 2 ,V a2
T i+4 :O
1, 000

phase [52].
(T i+4 , Zr+4 ) (O2 , V a2 )2
GZrO2 2
T i+4 :O
+GT IO2 + 35, 000
2
GZrO:O2
Zr +4
+GZRO2A + 5941 4.02T
GZrO2 a2
T i+4 :V
+GHSER + 200, 000

Ti
2
GZrO:V a2
Zr+4
+GHSER + 200, 000

Zr
0
0
0
0
0
LZrO2 +4 :O2
T i+4 ,Zr
15, 000 + 15.01T
LZrO2 +4 :O2
T i+4 ,Zr
+23, 000
LZrO2 +4 :O2
T i+4 ,Zr
2, 000 10T
2
LZrO:O2 ,V a2
Zr+4
130, 210 + 35.03T
2
LZrO:O2 ,V a2
Zr+4
19, 990

(rutile) phase [52].
(T i+4 , Zr+4 ) (O2 , V a2 )2
GZrO2 2
T i+4 :O
+GT IO2 + 35, 000
2
GZrO:O2
Zr+4
+GZRO2A + 11, 986 6.309T
0 ZrO2
GT i+4 :V a2
0 ZrO2
GZr+4 :V a2
0 ZrO2
LZr+4 :O2 ,V a2
+GHSER + 200, 000

Ti
0
0
+GHSER + 200, 000

Zr
235, 850 + 23T
218
(T i+4 , Zr+4 ) (O2 , V a2 )2

1
2
LZrO:O2 ,V a2
Zr +4
2
LZrO:O2 ,V a2
Zr +4
105, 440
+184, 910
Table C.14: Thermodynamic parameters for the 3 3

M6 X compound.
(T i)3 (Cu)2 (Cu, Al)1 (Cu, Al)1
0
6X
GMi:Cu:Cu:O
T
655, 832 + 141.67T
6X
GMi:Cu:Al:O
T
850, 050 + 221T
0
0
6X
LMi:Cu:Cu,Al:O
T
25, 000
Table C.15: Thermodynamic parameters for stoichiometric oxides.

Stoichiometric oxides
Cu O [71]
0
GCuO
Cu:O
+GCU O
0 Cu2 O
GCu:O
+GCU 2O
T i O [34]
GT iO
T i:O
GT IOA 12, 496 + 1.95751T
GT i2 O3
T i:O
GT I2O3 + 19508 2.4214T
GT i3 O2
T i:O
1, 064, 478 + 178.6T + 3 GHSER + 2 GHSER

Ti
O
GT i3 O5
T i:O
GT I3O5 + 15, 000 T
GT i4 O7
T i:O
GT I4O7 + 40T
Zr O [52]
GZr2 O
Zr:O
650, 000 + 145T + 2 GHSER + GHSER

Zr
O
GZr3 O
T i:O
HSER
605, 000 + 97T + 3 GHSER + GO
Zr
219
Stoichiometric oxides
0
GZr6 O
T i:O
594, 500 + 110.1T + 6 GHSER + GHSER

Zr
O
T i Zr O
2O
GZrT ii:O6
Zr:T
GZrO2A + 2 GT IO2 7250 + 25T
GZrT iO4
Zr:T i:O
GZrO2A + GT IO2 + 38493 20T
Cu T i O
0
GT i4 Cu2 O
T i:Cu:O
631, 814 + 116.51T + 4 GHSER + 2 GHSER + GHSER

Ti
Cu
O
Table C.16: Functions used in the models.

Name
Function
Range, K
Cu O functions [71]
GCU2OLIQ
GCUO
47734 + 148.5T 28T ln (T )
N/A
172, 735 + 291.78T 49.03T ln (T )
N/A
3.47 103 T 2 +
GCU2O
390,000
T
193, 230 + 360.057T 66.26T ln (T )

7.96 103 T 2 +
N/A
374,000
T
T i O functions [34]
GT IOA
551, 056.8 + 252.17T 41.99T ln (T ) 8.898 103 T 2

+1.097 108 T 3 +
327,015
T
1, 581, 243 + 940.17T 147.67T ln (T ) 1.74 103 T 2

GTI2O3
1.53 1010 T 3 +
N/A
470-2,115
2,395,423.5
T
159, 0717.7 + 1012.15T 156.9T ln (T )
2,115-3500
966, 880.6 + 348.55T 57.02T ln (T ) 2.017 102 T 2
300-700
+3.86 106 T 3 +
GTIO2
528,343
T
974, 253 + 461.2T 74.519T ln (T ) 1.356 103 T 2

+2.102 108 T 3 +
1,126,927
T
1, 022, 606 + 679.833T 100.42T ln (T )

220
700-2,130
2,130-4,000
Name
Range, K
2, 514, 503 + 1056.2T 174.75T ln (T ) 1.682 102 T 2
450-900
1, 772, 107 6, 305.1T + 882.36T ln (T ) 6.808 101 T 2

GTI3O5
Function
900-1,200
+7.767 105 T 3
96,994,950
T
2, 514, 675.5 + 1056.66T 174.79T ln (T ) 1.682 102 T 2
1,200-2,047
2, 566, 064.8 + 1501.09T 234.304T ln (T )
2,047-3,000
3, 510, 274 + 1643.59T 267.04T ln (T ) 2.467 102 T 2
500-1,000
+2.377 106 T 3 +
GTI4O7
3,660,393.5
T
3, 523, 423.8 + 1791.65T 288.799T ln (T ) 8.51 103 T 2

+1.118704 107 T 3 +
1,000-1,950
5,171,905
T
3, 642, 234.5 + 2439.22T 368.19T ln (T )
1,950-4,000
Zr O functions [52]
1, 125, 582 + 417.51T 68.328T ln (T ) 4.542 103 T 2
+1.724 1010 T 3 +
671,863
T
1, 129, 291 + 484.94T 78.1T ln (T ) 1.0845 1016 T 2

GZRO2A
+7.815 1021 T 3
221
2,620-2983
5.1278105
T
1, 193, 929 + 681.804T 100T ln (T ) + 3.041 1016 T 2

8.75 1021 T 3 +
1,445-2620
7.623108
1, 134, 269 + 501.79T 80T ln (T ) 2.0127 1014 T 2

+9.192 1019 T 3
298-1,445
2.467106
T
2,983-6,000
222
Appendix D
Qualitative Assessment of Diusion
Paths in AgCu-Ti Couples
The Ag Cu T i System is of great importance in the development of ceramic/metal joining techniques. This brief work has the intention of illustrating
how simple thermodynamic calculations and reasonable assumptions regarding the
kinetic behavior of the system can shed some light on the diusion paths followed by
diusion couples involving systems pertaining to this ternary once particular initial
conditions are established.
D.1
Estimation of Diusion Paths
When two dierent phases of dierent compositions and brought into contact, the
dierent chemical potentials of the elements comprising the two phases constitute the
driving force for mass diusion. The components of each phase would follow a diusion path determined by the experimental conditions and the actual thermochemical
properties of the system. In theory, in a system of third and higher order, the diusion path actually followed by a diusion couple constructed from two distinct phases
within the system could not be determined without any knowledge on the actual kinetic properties of the phases in question. The reason for this is the innite number
of trajectories through which a system can achieve thermodynamic equilibrium.
Despite this, thermodynamic calculations can be of use in determining possible
diusion paths if some simple and reasonable assumptions are made, as has been
shown by [120]. One can assume, for example that there exist local equilibrium conditions at each one of the interfaces. If this holds true, it is reasonable to calculate
metastable equilibrium conditions and assume that it is possible to determine which
phases are formed by simply calculating the driving forces for precipitation of the
remaining phases in the system. To nd the actual metastable equilibrium state,
inferences regarding the kinetics of the system could be made. For example, if the
diusion rates in both phases are very dierent from each other reasonable assumptions regarding the ratio of the components concentration proles can be made.
223
With these simple assumptions, a possible metastable equilibrium state can be

calculated. At this particular point in thermodynamic space it is possible to calculate
the driving force for precipitation of the rest of the phases not playing a role in the
mestastable equilibrium:
G (J/mol)
Phase
Phase
Phase
Phase
DGMR()
DGMR()
A
Figure D-1: Schematics of Metastable Equilibrium.
Fig. D-1 illustrates a mestastable equilibrium condition between two phases, while
two other phases have positive driving forces for precipitation but are limited by kinetics. The initial condition is a diusion couple between phases and . At the
beginning of this reaction, a mestastable equilibrium between these two phases is
established. Although phases and are thermodynamically stable, energy barriers due to surface energy or mechanical strains may prevent them from nucleating.
To determine which of the two dormant phases precipitates rst, we could use
elementary nucleation theory:
J = f (T ) e
16 3
3(Gv Gs )2
(D.1)
where J is the nucleation rate; is the interfacial energy between the nucleus and
the matrix;Gv corresponds to the driving force of precipitation of the phase; and
Gs is the elastic strain energy contribution to the volumetric energy.
As can be seen, to estimate the relative nucleation rates for competing phases
and , it would be necessary to have information regarding interfacial and elastic
strain energies which, due to their nature, are extremely dicult to obtain experimentally. However, it is reasonable to assume that interfacial and strain energies would
be in the same order of magnitude for each of the phases in question. Therefore,
the dierence in the value for the driving force for precipitation would then be the
determining factor to establish which of the two dormant phases would precipitate
rst, under the particular metastable equilibrium conditions [120].
224
D.2
Application to the AgCu-Ti Diusion Couple
In their experiments regarding the Ag Cu T i System, Paulasto et al. [22] studied

the thermochemical interactions between several liquid AgCu alloys and T i. Due to
the reactivity of T i and the high vapor pressure of Ag, the AgCu alloys were placed
inside T i ampules, which were later evacuated and sealed. The samples were heated to
9000 C, 9500 C and 10000 C for several periods of time and were later water-quenched.
Based on the microstructural and compositional analysis of the phases present, they
were able to conrm the main features of the Ag Cu T i system, as described
by Eremenko et al. [41, 42, 43, 44]. Using the approach outlined above, it is possible
to rationalize the ndings reported in the work by Paulasto et al. [22] and propose
a possible diusion-reaction path consistent with both the existing thermodynamic
information on the Ag CuT i (see Section 3.2) system and reasonable assumptions
regarding the kinetic behavior of the phases involved.
1.0
1.0
0.9
0.9
0.7
tio
n
0.6
rac
rac
tio
n
0.7
Mo
0.6
0.5
le F
le F
0.5
0.4
Mo
0.8
X(
Ti)
X(
Ti)
0.8
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0.2
0.4
0.6
0.8
Mole Fraction X(Cu)
1.0
1.0
0.2
0.4
0.6
0.8
Mole Fraction X(Cu)
A)
1.0
)
nX
(Ti
)
nX
(Ti
0.8
tio
0.6
Fra
c
tio
Fra
c
0.8
0.7
0.7
0.5
Mo
le
Mo
le
B)
BCC
0.9
0.9
0.4
0.6
CuTi
CuTi2
CuTi
0.4
Liq#1
+Liq#2
0.3
0.3
0.2
0.2
0.1
0.1
0
0
1.0
0.2
0.4
0.6
0.8
Mole Fraction X(Cu)
1.0
C)
Liquid
0.2
0.4
0.6
0.8
1.0
FCC
Mole Fraction X(Cu)
D)
Figure D-2: Proposed diusion paths and experimental results by Paulasto et al. [22].
Fig. D-2 A) shows the proposed initial conditions of the system: At the beginning
of the reaction, T i is in contact with a liquid AgCu eutectic alloy. The eutectic
One eutectic and three non-eutectic alloys.
225
composition is approximately 60 %Ag 40 %Cu. To study this initial condition, a

mestastable equilibrium ternary diagram involving the liquid and the solid bcc phases
is calculated. When the interfacial reaction starts, T i would rapidly diuse out and
dissolve into the liquid continuously until the liquid is saturated with the element .
During the initial stages of this interfacial reaction, the concentration of T i at the interface can be assumed to lie somewhere at the (bcc+liquid)/liquid phase boundary .
On the other hand, the elements in the liquid phase (Ag and Cu) will also start
do diuse into the bcc phase. Because of the expected dierence in the diusion rates
n(Ag)
of Ag and Cu in bcc T i, it would be expected that the ratio n(Cu) would be dierent
from the initial conditions. To examine this, it is important to consider the ratio of
the diusivities in the liquid and bcc phases. To a rst approximation, tracer diusiv2
ities in the liquid state can be reasonably assumed to be in the order of 1 102 m ,
s
2
while diusion coecients in the solid state would be in the order of 1 1012 m .
s
Concentration gradients in the liquid state would be negligible compared to those of
the solid phase and this fact could be use to estimate the proper concentration at the
liquid/(bcc + Liquid) phase boundary. A complication arises, however, due to the
existence of a miscibility gap in the liquid phase.
To establish a reasonable diusion path at the onset of the interfacial reaction,
the combined observations regarding the behavior of the liquid and solid phases could
be used:
x
Ag
Starting from the AgCu eutectic point, a line fullling the condition xCu = 1.5
could be traced for increasing amounts of T i, which would represent continuous T i
dissolution in the liquid phase. As can be seen from Fig. D-2 A), this line intersects
the miscibility gap boundary at around xT i = 0.1. Since the system must be at
equilibrium , we can assume that the interface between Liquid 1 and Liquid 2 is
dened by a tie-line, which implies a planar liquid/liquid interface.
To complete the analysis, the diusion processes occurring at the solid interface
must also be considered. The actual composition at the metastable Liquid 2/bcc
interface must be determined. It is reasonable to assume that the T i content at this
interface must be such that a high T i chemical potential is obtained, since, as the
reaction progresses, the chemical potential of T i at the interface is expected to decrease. The candidate composition at the bcc/(bcc + liquid 2) phase boundary would
then be located at the xCu 0.2, xT i 0.8 compositional coordinate. By assuming
again a planar liquid 2/bcc interface, it is possible to nd a tie-line that connects the
Including the well-established miscibility gap.

When the chemical potential of T i is the same in the liquid and the solid phases.
This assumption, however, is complicated by the existence of a miscibility gap in the liquid
phase.
Liquid 1 + Liquid 2
A simplifying assumption would be that the two liquid phases have the same density and
therefore the interfaces remain unchanged throughout the reaction
226
compositions at the bcc/(bcc + liquid 2) and (bcc + liquid 2)/liquid 2 phase boundaries. The rest of the proposed diusion path is shown in Fig. D-2 A). It is important
to note that the proposed diusion path crosses the dotted line connecting the AgCu
eutectic and the pure T i compositions. This is a necessary condition if mass is to be
conserved in the system..
To estimate which phase would precipitate next under the already dened metastable
conditions, it is necessary to calculate the driving force for precipitation for all the
stable phases. Fig. D-3 shows the calculated ratio of the driving forces for precipitation for the CuT i2 and CuT i phases along the bcc/(bcc + liquid 2) phase boundary as
a function of T i composition. The xT i 0.8 corresponds to the xCu 0.2, xT i 0.8
compositional coordinate. As noted above, this point has the highest T i chemical
potential and it seems reasonable to consider that it represents the initial interfacial
condition at the onset of the reaction. Fig. D-3 shows that at this composition, the
driving force for precipitation of the CuT i2 is ve times as large as that of the CuT i
phase. It is therefore reasonable to expect that this phase would be the rst to form
under these diusion/reaction conditions.
Fig. D-2 B) shows a diusion path at t = t1 > t0 . By this time, the CuT i2
phase is assumed to have already precipitated. Since the expected Cu content of this
phase at this time is higher than that of Ag, the liquid phase would be depleted of
xAg
Cu and the xCu ratio would have increased. This is shown by decreasing the slope
of the diusion path connecting the AgCu eutectic point and the liquid 1 + liquid 2
phase boundary. At the same time, the T i content at the bcc side of the bcc/liquid 2
interface would have decreased, due to the formation of the CuT i2 particles and the
continuing dissolution of T i in the liquid phase, accompanied by the enrichment of the
bcc phase with Ag and Cu. The tie-line connecting the two liquid phases corresponds
to a higher T i chemical potential, as expected due to the further T i enrichment of
the liquid.
Fig. D-2 C) shows yet another diusion path at t = t2 > t1 > t0 . It is assumed
that the CuT i phase has already precipitated and therefore, a metastable equilibrium
between T i, CuT i2 , CuT i, Liquid 1 and Liquid 2 has to be calculated. The composition of the CuT i phase is richer in Cu than in Ag and a further decrease in the
slope connecting the AgCu eutectic point and the Liquid 1 + Liquid 2 phase boundary is expected. As the chemical potential of T i at the T i side of the T i/Liquid 2
interface decreases, the diusion path indicates that the CuT i2 phase would increase
its Ag content. The tie-line connecting the two liquid phases would correspond to a
higher T i chemical potential, due to further dissolution of T i into the liquid.
Fig. D-2 D) shows the experimental compositions at 9500 C after annealing for
30 min.. As can be seen, the compositions observed are consistent with the ternary
phase diagram. Moreover, the experimental compositions are consistent with the diffusion path proposed in this work. The validity of the analysis and the assumptions
made are thus justied. A similar approach can be used to analyze diusion reactions
227
DGMR(CuTi2)/DGMR(CUTI)
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.70
0.72
0.74
0.76
X(TI)
0.78
0.80
Figure D-3: DGM R(CuT i2 ratio at the T i/( T i + Liquid 2) phase boundary as a function
DGM R(CuT i)
of T i content.
of this system at other temperatures of interest.
D.3
Summary
In this appendix, following a similar approach by Lee et al. [120] it has been demonstrated the usefulness of metastable thermodynamic calculations to establish possible
diusion paths in diusion couples in ternary systems. Diusion experiments on the
Ag Cu T i system have been rationalized using the complete thermodynamic
description of the Ag Cu T i system.
228
Appendix E
Numerical Implementation of
Phase Field Model of Coupled
Oxide Layers Growth
E.1
Finite Dierence Formulation
To solve the Cahn-Hilliard equation (Eq. 7.13) for this system, a nite dierence formulation was used. For simplicity, a one-dimensional problem was considered. For
the rst term of the C-H equation, three uniformly spaced grid points (h spacing)
were used. The discretization error of this nite dierence formulation is O(h2 ):
M
M
f
c
( (O)) = M
( (O))
i1 2i +i+1
h2
For the second term, ve equally spaced grid points were used and a central fourth
order dierence equation of discretization error O(h2 ) was obtained:
2k 2 c(x) = M
2k 2 x(O)
x(O)i2 4x(O)i1 + 6x(O)i 4x(O)i+1 + x(O)i+2
4
x(O) = 2kM
h4
M
2kM
E.2
Treatment of Boundary Conditions
To simplify the implementation of the numerical simulation, central dierences were

used also at both boundaries. The system was treated as if it had symmetric boundary conditions. The computational domain to the right is the mirror image to the
one on the left:
Therefore, special nite dierence equations were implemented at the boundaries.
The code was written so the boundary conditions could be easily modied in case
dierent physical situations were to be simulated. As can be seen, in the simplied
model, the zirconia is not considered and instead, a concentration is xed at one end
229
J=0
x(O)=Constant
J=0
Figure E-1: Boundary conditions for simulations.
of the computational domain. Although this is not entirely correct, it still resembles
the physics of the original problem.
E.3
Value of the gradient energy term
In the diuse-interface formalism, one of the key parameters determining the interface
characteristic thickness is the gradient energy coecient (See Eq. 7.13), k. As described in Section 7.4, the gradient energy coecient is related to the energy penalty
that the system has to pay in order to have abrupt changes in composition (i.e. interfaces). The larger this parameter is, the greater the importance of the second
parameter in Eq. 7.13. It is therefore important to be able to assign a value for this
2
parameter. It is known (See Section 7.4) that k 2 , where is the interface
thickness. By assuming that:
=3h
(E.1)
k can be obtained through
k=K
3
h
2
(E.2)
where K is a constant that reects the uncertainties in the value for the parameter k.
It is obvious that, given the actual widths of real interfaces ,the parameter K must be
modied in order to make the calculations practical. Although the actual interface is
not accurately described (its dimension is increased by several orders of magnitude),
the essential aspects of the diuse-interface approached are retained through the approximation shown in Eq. E.2.
E.4
Numerical Scheme: Explicit Euler
The simplest way to numerically integrate the marching time Cahn-Hilliard equation
is to implement a simple forward Euler numerical scheme in time:
x(O)n+1 = x(O)n + t M D2 ((O)n ) 2 D4 (x(O)n )
i
i
i
i
(E.3)
where t is the time interval, n denotes the previous time interval, n+1 corresponds
to the interval to calculate, D2 is the discrete second spatial derivative operator and
A typical value is about 10 monatomic layers, or 30 .

A
230
D4 is the fourth order nite dierence operator.

Since this numerical scheme is conditionally stable, there is a maximum allowable
t, beyond which the solution becomes unstable and the simulation diverges. An
additional problem is the fact that at the beginning of the simulation, the system develops the interfaces quite rapidly and as the simulation proceeds, the evolution of the
system becomes increasingly slower. This means that there are dierent time scales
during the simulation and thus an adaptive time stepping scheme was implemented.
To ensure stability of the simulation, the maximum change in the concentration was
xed.
The algorithm is designed so the t used in the integration reaches a steady state
and the simulation remains stable. Basically, three dierent time scales in the simulation can be observed. The rst one involves the initial formation of the two diuse
interfaces. At this stage, large concentration gradients develop and large local uxes
are present. As the simulation continues, Dt increases at a slower pace. Eventually,
the simulation reaches a stage in which local variations in the composition prole
occur slowly. The large variation at some stages of the simulation can be minimized
by decreasing the tolerance.
E.5
Numerical Scheme: Semi-implicit Crank-Nicholson
For the conventional diusion equation, semi-implicit schemes such as C-N are unconditionally stable. Unfortunately, this is not true for the C-H equation, because
of the non-linearity and the bi-harmonic term . Despite this, it is expected that the
maximum allowable Dt is greater than that of the simple explicit scheme. It will be
shown that this is indeed the case. Furthermore, it will be proven that as the required
accuracy of the numerical simulation is increased, the use of semi-implicit schemes is
increasingly more advantageous.
The C-N numerical scheme can be represented by:
x(O)n+1
i
x(O)n + t
i
[M D2 ((O)n ) 2 D4 (x(O)n )]
i
i
+ M D2 (O)n+1 2 D4 x(O)n+1
i
i
2
(E.4)
In this method, the PDE is supposed to be satised at a mid-point between the

actual time and the immediate next time in the future. Therefore, the solution in
time is related to the average to the spatial solutions in both the actual time and the
immediately next one.
231
E.6
Newton Method
As can be seen in Eq. E.4, the C-N method involves the solution of a non-linear system of equations. Because of the non-linearity of the system of equations, an iterative
method must be used. For this particular problem, the multi-dimensional Newton
method was used.
0.8
f(x0)+f(x 0)(x0 x1)=0

0.6
0.4
f(x)
0.2
x2
x1
x0
0. 2
0. 4
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure E-2: One-dimensional Newton method.
Fig. E-2 shows the concept of the Newton method applied to a simple onedimensional problem. This method relies on the use of the local derivative of the
function evaluated at the initial guess (X0 ) to approach the solution (X1 ) . The procedure continues until the solution has been reached (converged) within a pre-dened
tolerance (X2 ). In the case of the multidimensional Newton method, the algorithm
would be:
To solve F (x) where x is a vector to obtain xn+1 , do:
1) x0 = initial guess, k=0
2) Repeat {
3) Compute F (xk ), JF (xk ) (Jacobian Matrix)
4) Solve JF (xk ) (xk+1 xk ) = F (xk ) for xk+1
5) k = n + 1
6) } Until xk+1 xk AND f xk+1
are small enough and where JF (xk ) is the

232
Jacobian matrix and is dened as:
..
.
F1 (x)
cN
FN (x)
x1
JF xn , xk
F1 (x)
x1
FN (c)
xN
.
.
.
.
.
.
x1
.
.
.
xN
For the problem at hand:
F (xi ) =
xn+1
i
xn
i
+ t
[M D2 (n ) 2 D4 (xn )]
i
i
+ M D2 n+1 2 D4 xn+1
i
i
2
(E.5)
One of the main diculties in implementing Newton iterative methods is the calculation of the Jacobian matrix. The algebraic expression describing how the chemical
potential, , changes with composition will have to be used and the solver would
have to be modied each time this expression changes. Thus, if a general C-H semiimplicit solver is to be implemented, other ways for approximating the Jacobian must
be found:
For this work, a Jacobian-free Newton iteration method [171, 172] was used. The
Jacobian was approximated using a Gradient Conjugate Residual GCR algorithm.
This algorithm attempts to minimize
rk+1 = F xk JF xk xk+1 xk
(E.6)
At each iteration of the GCR, JF (xk ) can be approximated by

1
F xk + p F xk
(E.7)
where p is an orthogonalized search direction for the particular GCR iteration.

J F xk p =
E.7
Implementation of Semi-Implicit Numerical Scheme
For this implementation, the dierent time-scales of the simulation were also considered and an adaptive time-stepping scheme was implemented. This adaptive timestepping was designed such that the simulation would minimize the number of Newton
iterations required to solve each time interval and at the same time it would maximize
the integrating t.
233
234
Appendix F
Experimental Apparatus
Reactive brazing involves the use of highly reactive elements that are susceptible
to oxidation. For reactive brazing of zirconia to be successful, it is necessary to ensure
that the only possible source of oxygen (for the oxidation of the reactive element) is
the zirconia itself and not the atmosphere surrounding the specimen.
Figure F-1: Experimental apparatus.
Fig. F-1 presents a schematics of the experimental apparatus designed for the
experimental part of this work. With this experimental apparatus it was possible
to control the thermal history of the samples studied, while maintaining the oxygen
partial pressure low enough so no reaction between the active alloys and the atmosphere would take place. The compressive on the load was controlled through a load
gauge that was positioned at one of the ends of the sample holder. With this furnace,
235
vacuum pressures of 106 torr were routinely obtained. The maximum operating temperature of this furnace was 12500 C. An actual picture of the experimental setup is
presented in Fig. F-2
Figure F-2: Vacuum high-temperature furnace.
Arroyave [51] provides a more detailed description of the characteristics and capabilities of the experimental apparatus presented in this appendix.
236
Bibliography
[1] M. G. Nicholas. Ceramic-Metal Interfaces. In Surfaces and Interfaces of Ceramic Materials, pages 393417, United Kingdom, 1989. Kluwer Academic Publishers.
[2] J.-W. Park, P. F. Mendez, and T. W. Eagar. Strain Enegy Distribution in
Ceramic/Metal Joints. Acta Mater., 50:883899, 2002.
[3] J. M. Howe. Bonding, Structure and Properties of Metal/Ceramic Interfaces:
Part 1 Chemical Bonding, Chemical Reaction and Interfacial Structure. Int.
Mater. Rev., 38:233255, 1993.
[4] A. Dupr. Thorie Mcanique de la Chaleur, chapter IX. Gauthier-Villars,
e
e
e
Paris, 1869.
[5] N. Eustathopoulos, M. G. Nicholas, and B. Drevet. Wettability at High Temperatures. Pergamon, 1999.
[6] J. E. MacDonald and J. G. Eberhart. Adhesion in Aluminum Oxide-Metal
Systems. Trans. Metall. Soc. AIME, 233:512517, 1965.
[7] B. M. Gallois. Wetting in Non-reactive Liquid-Metal Systems. JOM, 49(6):
4851, 1997.
[8] M. W. Chase. JANAF Thermochemical Tables. American Chemical Society,
Washington DC, 1983.
[9] J.-G. Li. Wetting and Interfacial Bonding of Metals with Ioncovalent Oxides.
J. Am. Ceram. Soc., 75:31183126, 1992.
[10] F. R. de Boer, R. Boom, W. C. M. Mattens, A. R. Miedema, and A. K. Niessen.
Cohesion in Metals, Transition Metal Alloys. North-Holland, Amsterdam, 1988.
[11] I. Rivollet. Contact Angle and Thermodynamic Adhesion in Nonreactive MetalAlumina Systems. PhD thesis, Institut National Polytechnique, 1986.
[12] W. H. Strehlow and E. L. Cook. Compilation of Energy Band Gaps in Element
and Binary Compound Semiconductors and Insulators. J. Phys. Chem. Ref.
Data, 2:16393, 1973.
237
[13] O. Kubaschewski, C. B. Alcock, and P. J. Spencer. Materials Thermochemistry.

Pergamon Press, Oxford, sixth edition, 1993.
[14] R. M. Pilliar and J. Nutting. Solid-Solid Interfacial Energy Determinations in
Metal-Ceramic Systems. Philos. Mag., 16:181188, 1967.
[15] M. G. Nicholas. Active Metal Brazing. Br. Ceram. Trans. and J., 85:144146,
1986.
[16] N. Eustathopoulos and B. Drevet. Mechanisms of Wetting in Reactive
Metal/Oxide Systems. Mater. Res. Soc., 314:1525, 1993.
[17] N. Eustathopoulos, M. G. Nicholas, and B. Drevet. Wetting at High Temperatures, chapter 2, pages 9192. Pergamon, 1999.
[18] F. Moret and Eustathopoulos. Ceramic to Metal Direct Brazing. J. de Physique
IV, 3:10431052, 1993.
[19] C. H. Lupis. Chemical Thermodynamics of Materials. Prentice Hall, New York,
1993.
[20] Y. Naidich. The Wettability of Solids by Liquid Metals. In D. A. Cadenhead
and J. F. Danielli, editors, Progress in Surface and Membrane Science, page
353, New York, 1981. Academic Press.
[21] N. Saunders and A. P. Miodownik. CALPHAD - A Comprehensive Guide.
Pergamon, New York, 1998.
[22] M. Paulasto, F. J. J. Van Loo, and J. K. Kivilahti. Thermodynamic and experimental study of Ti-Ag-Cu alloys. J. Alloy. Compd, 220:136141, 1995.
[23] G. P. Kelkar and A. H. Carim. Synthesis, Properties, and Ternary Phase Stability of M6 X Compounds in the Ti-Cu-O System. J. Am. Ceram. Soc., 76:
18151820, 1993.
[24] O Botstein, A. Schwarzman, and A. Rabinkin. Induction Brazing of Ti-6Al-4V
Alloy with Amorphous 25Ti-25Zr-50Cu Brazing Filler Metal. Mater. Sci. Eng.
A, A206:1423, 1996.
[25] Z. S. Yu, M. F. Wu, C. Liang, and K. Wang, F. J. Amd Qi. Interfacial Morphology and Strength of Al2 O3 /Nb Joint Brazed with Cu-Ti-Zr Filler Metal.
Mater. Sci. Technol., 18:99102, 2002.
[26] R. Arroyave and T. W. Eagar. Thermodynamic Assessment of the System
Cu-Ti-Zr. J. Alloys Compd., 351:158170, 2002.
[27] R. Arroyave. Assessment of the Ag-Ti System. Technical report, Massachusetts
Institute of Technology, 2002.
238
[28] J. J. Pak, M. L. Santella, and R. J. Fruehan. Thermodynamics of Ti in Ag-Cu

Alloys. Metall. Trans. B., 21B:349355, 1990.
[29] F. H. Hayes, L. Lukas, G Eenberg, and G. Petzow. A Thermodynamic Optimization of the Cu-Ag-Pb System. Z. Metallkd., 77:749754, 1986.
[30] K. C. Hari Kumar, I. Ansara, P. Wollants, and L. Delaey. Thermodynamic
Optimization of the Cu-Ti System. Z. Metallkd., 87:666672, 1996.
[31] J. L. Murray and K. J. Bhansali. The Ag-Ti (Silver-Titanium) System. Bull.
Alloy Phase Diagrams, 4:178182, 1983.
[32] K. Fitzner and O. J. Kleppa. Thermochemistry of Binary Alloys of Transition
Metals: The Me-Ti, Me-Zr, and Me-Hf (Me=Ag,Au) Systems. Metall. Mater.
Trans. A, 23A:9971003, 1992.
[33] P. Wei, L. Rongti, C. Jian, S. Ruifeng, and L. Jie. Thermodynamic Properties
of Ti in Ag-Ti Alloys. Mater. Sci. Eng. A, A287:7277, 2000.
[34] B. J. Lee. Thermodynamic Evaluation of the T i O Binary System. J. Korean
Inst. Metals and Materials, 32:869877, 1994.
[35] C. Wagner. Thermodynamics of Alloys, chapter 2, pages 5458. AddisonWesley, 1952.
[36] L. Kaufman. Thermodynamic Database for the Ag-Ti System, 2002. Personal
Communication, MIT, 2002.
[37] M. R. Plichta and H. I. Aaronson. The Transformation in Three Ti-X
Systems. Acta Metall., 26:12931305, 1978.
[38] R. Reinbach and D. Fischmann. Diusion in the Titanium-Silver System. Z.
Metallkd., 54:314316, 1963.
[39] M. K. McQuillan. A Study of the Titanium-Silver System. J. Inst. Met., 88:
235239, 1960.
[40] H. Bakker. Enthalpies in Alloys: Miedemas Semi-Empirical Model. Trans Tech
Publications Ltd, Zurich, 1998.
[41] V. N. Eremenko, Y. I. Buyanov, and N. M. Panchenko. Phase Separation in
the Molten State in the Ti-Cu-Ag System. Izv. Akad. Nauk SSSR, Metally, 5:
200202, 1969.
[42] K. Eremenko, Yu. I. Buyanov, and N. M. Panchenko. Phase Equilibrium in
the Ti-Cu-Ag System at 7000 C. Izv. Akad. Nauk. USSSR, Metally, 3:188193,
1969.
239
[43] N. M. Eremenko, Yu. I. Buyanov, and N. M. Panchenko. The Liquidus Surface

of the System Titanium-Copper-Silver. Sov. Powder Metall. Met. Ceram., 10:
301304, 1970.
[44] N. M. Eremenko, Yu. I. Buyanov, and N. M. Pancheno. Polythermal and
Isothermal Sections of the System Titanium-Copper-Silver. Sov. Powder Metall.
Met. Ceram., 10:410414, 1970.
[45] C. G. Woychik and T. B. Massalski. Phase Diagram Relationships in the System
Cu-Ti-Zr. Z. Metallkd., 79:149153, 1988.
[46] W. D. MacDonald and T. W. Eagar. Transient-liquid-phase Bonding. Annu.
Rev. Mater. Sci., 22:2346, 1992.
[47] V. N. Chebotnikov and V. V. Molokanov. Structures and Properties of Amorphous and Crystalline Alloys in the T i2 Cu-Zr2 Cu Section in the Cu-Ti-Zr System. Inorg. Mater., 26:808811, 1990.
[48] K. J. Zeng, H al
am ainen, and H. L. Lukas. A New Thermodynamic Description
of the Cu-Zr System. J. Phase Equilib., 15:577586, 1994.
[49] C. C. H. Kumar, P. Wollants, and L. Delaey. Thermodynamic Assessment of
the Ti-Zr System and Calculation of the Nb-Ti-Zr Phase Diagram. J. Alloys
Compd., 206:121127, 1994.
[50] BJ. Lee. Prediction of T i/Al2 O3 Interface Reaction Products by Diusion Simulation. Acta. Mater., 45:39933999, 1997.
[51] Raymundo Arroyave. Reactive Brazing of Zirconia to Nickel-Based Super Alloys. Masters thesis, Massachusetts Institute of Technology, 2000.
[52] R. Arroyave, L. Kaufman, and T. W. Eagar. Thermodynamic Modelling of the
Zr-O System. CALPHAD, 26:95118, 2002.
[53] L. Kaufman and E. V. Clougherty. Thermodynamic Factors Controlling the
Stability of Solid Phases at High Temperatures and Pressures. In AIME Metall.
Soc., editor, Metallurgy at High Pressures and High Temperatures, pages 322
379, New York, 1964. Gordon and Breach, Science Publishers.
[54] J. P. Abriata and R. Versaci. The O-Zr (Oxygen-Zirconium) System. Bull. of
Alloy Phase Diagrams, 7(2):116124, 1986.
[55] M. Hillert, B. Jansson, B. Sundman, and Agren. A Two-Sublattice Model for

Molten Solutions With Dierent Tendency for Ionization. Metall. Mater. Trans.
A, 16A:261, 1985.
[56] T. Tsuji. Thermochemistry of IVA Transition Metal-Oxygen Solid Solutions.
J. Nucl. Mater., 247:6371, 1997.
240
[57] P. Waldner and G. Eriksson. Thermodynamic Modelling of the System

Titanium-Oxygen. CALPHAD, 23:189218, 1999.
[58] M. Hirabayashi, S. Yamaguci, H. Asano, and K. Hiraga. Order-Disorder transformations of Interstitial Solutes in Transition Metals of IV and V Groups. In
H. Warlimont, editor, Order-Disorder Transformations in Alloys, pages 266
302, Berlin, 1974.
[59] R. W. Vest, N. M. Tallan, and W. C. Tripp. Electrical Properties and Defect
Structure of Zirconia: I, Monoclinic Phase. J. Am. Ceram. Soc., 47:635640,
1964.
[60] R. W. Vest and N. M. Tallan. Electrical Properties and Defect Structure of
Zirconia: II, Tetragonal Phase and Inversion. J. Am. Ceram. Soc., 48:472475,
1965.
[61] M. Hillert and B. Jansson. Thermodynamic Model for Non-stoichiometric Ionic
Phases. Application to CeO2x . J. Am. Ceram. Soc., 69:732734, 1986.
[62] S. C. Carniglia, S. D. Brown, and T. F. Schroeder. Phase Equilbria and Physical
Properties of Oxygen-Decient Zirconia and Thoria. J. Am. Ceram. Soc., 54:
1317, 1970.
[63] R. J. Ackerman, S. P. Garg, and E. G. Rauh. High-Temperature Phase diagram
for the System Zr-O. J. Am. Ceram. Soc., 60:341345, 1977.
[64] R. J. Ackerman, S. P. Garg, and E. G. Rauh. The Lower Phase Boundary of
ZrO2x . J. Am. Ceram. Soc., 61:275276, 1978.
[65] E. G. Rauh and P. Garg. The ZrO2x (cubic)- ZrO2x (cubic+tetragonal)
Phase Boundary. J. Am. Ceram. Soc., 63:239240, 1980.
[66] E. Gebhardt, H. D. Seghezzi, and W. Durrschenabel. Research on the System
Zirconium-Oxygen. J. Nucl. Mater., 4:255268, 1961.
[67] G. Boureau and P. Gerdanian. Use of a Tian-Calvet Microcalorimeter at 1300o C
Direct Measurement of hM2 in the Metal-Oxygen Systems. Can. Metall. Q., 13:
O
339343, 1974.
[68] G. Boureau and P. Gerdanian. High Temperature Thermodynamics of Solutions
of Oxygen in Zirconium and Hafnium. J. Phys. Chem. Solids, 45:141145, 1984.
[69] K. L. Komarek and M. Silver. Thermodynamic Properties of ZirconiumOxygen, Titanium-Oxygen and Hafnium-Oxygen Alloys. In Thermodynamics of
Nuclear Materials, pages 749774, Vienna, Austria, 1962. International Atomic
Energy Agency.
[70] R. J. Ackerman, S. P. Garg, and E. G. Rauh. The Thermodynamic Properties
of Substoichiometric Zirconium Dioxide at the Lower Phase Boundary. High.
Temp. Sci., 11:199210, 1979.
241
[71] B. Hallstedt, D. Risold, and L. J. Gauckler. Thermodynamic Assessment of the

Copper-Oxygen System. J. Phase Equilib., 15:483499, 1994.
[72] G. P. Kelkar, K. E. Spear, and Carim A. H. Thermodynamic Evaluation of
Reaction Products and Layering in Brazed Alumina Joints. J. Mater. Res., 9:
22442250, 1994.
[73] M. H. Mueller and H. W. Knott. The Crystal Structures of T i2 Cu, T i2 N i,
T i4 N i2 O and T i4 Cu2 O. Trans. Metall. Soc. AIME, 227:67478, 1963.
[74] G. P. Kelkar and A. H. Carim. Al solubility in M6 X Compounds in the Ti-Cu-O
System. Mat. Letters, 23:231235, 1995.
[75] C. Peytour, F. Barbier, P. Berhet, and A. Revcolevschi. Characterization of
Al2 O3 /T A6V and ZrO2 /T A6V Ceramic-Metal Interfaces. J. de Phys., 51,
Suppl.C1:C18 97 C19 02, 1990.
[76] M. L. Santella, J. A. Horton, and J. J. Pak. Microstructure of Alumina Brazed
with a Silver-Copper-Titanium Alloy. J. Am. Ceram. Soc., 73:17851787, 1990.
[77] H. Haessler, H. Kippenberg, and A. G. Siemens. Possibilities of degassing alloys
based on copper with elements having an oxygen anity such as chromium,
zirconium, titanium in relation to the material of the crucible. In J. G. Krueger,
editor, Proc. Int. Conf. Vac. Metall. Electroslag Remelting Processes, pages 87
89, Hanau, Germany, 1977. Leybold-Heraeus GmbH Co.
[78] V. S. Sudavtsova. Thermodynamic Properties of the Cu-O-Zr System. J.
Ukrainian Chem., 59:11491150, 1993.
[79] C. H. Lupis. Chemical Thermodynamics of Materials. Prentice-Hall, New York,
1983.
[80] B. Sundman. Modication of the Two-Sublattice Model for Liquids. CALPHAD, 15:109119, 1991.
[81] J. Assal, B. Hallstedt, and L. J. Gauckler. Thermodynamic Assessment of the
Ag-Cu-O System. J. Phase Equilib., 19:351361, 1998.
[82] C. Wagner. The Activity Coecient of Oxygen and Other Nonmetallic Elements
in Binary Liquid Alloys as a Function of Alloy Composition. Acta Metall., 21:
12971303, 1973.
[83] H. M. Ondik and H. F. (Editors) McMurdie. Phase Diagrams for Zirconium
and Zirconia Systems. The American Ceramic Society, Westerville, Ohio, USA,
1998.
[84] J.-H. Park, P. Liang, H. J. Seifert, F. Aldinger, B.-K. Koo, and H.-G. Kim.
Thermodynamic Assessment of the ZrO2 and T iO2 System. Kor. J. Ceram.,
7:1115, 2001.
242
[85] F. H. Brown and P. Duwez. The Zirconia-Titania System. J. Am. Ceram. Soc.,
37:129132, 1954.
[86] A. H. Webster, R. C. MacDonald, and W. S. Bowman. The System P bO-ZrO2 T iO2 at 1100. J. Canad. Ceram. Soc., 34:97102, 1965.
[87] M. Hoch and R. L. Dean. The System Titanium-Zirconium-Oxygen. Trans.
Metall. Soc. AIME, 221:11621173, 1961.
[88] A. V. Shevchenko, L. M. Lopato, I. M. Maister, and O. S. Gorbunov. The
T iO2 ZrO2 System. Russ. J. Inorg. Chem., 25:13791381, 1980.
[89] T. Noguchi and M. Mizuno. Phase Changes in Solids Measured in Solar Furnace,
ZrO2 -T iO2 System. Solar Energy, 11:5661, 1967.
[90] B. Jansson. Evaluation of Parameters in Thermochemical Models using different types of Conditions. Technical Report TRITA-MAC-0234, Division of
Metallurgy, Royal Institute of Technology, Stockholm, Sweden, 1984.
[91] L. W. Coughanour, R. S. Roth, and V. A. DeProsse. Phase Equilibrium Relations in the Systems Lime-Titania and Zirconia-Titania. J. Res. Natl. Bur.
Std., 52:3742, 1954.
[92] T. Noguchi and M. Mizuno. Phase Changes in the ZrO2 T iO2 System. Bull.
Ceram. Soc. (Jap.), 41:28952899, 1968.
[93] A. E. McHale and R. S. Roth. Low Temperature Phase Relationships in the
System ZrO2 -T iO2 . J. Am. Ceram. Soc., 69:827832, 1986.
[94] W. Weppner. Tetragonal Zirconia Polycrystals- a High Performance Solid Oxygen Conductor. Solid State Ionics, 52:1521, 1992.
[95] R. V Allen, W. E. Borbidge, and P. T. Whelan. The Reaction-Bonded Zirconia Oxygen Sensor: An Application for Solid-State Metal-Ceramic ReactionBonding. Advances in Ceram., 12:53743, 1984.
[96] W. B. Hanson, K. I. Ironside, and J. A. Fernie. Active Metal Brazing of Zirconia.
Acta Mater., 48:46734676, 2000.
[97] R. C. Garvie. Structural Applications of Zirconia-bearing Materials. Advances
in Ceramics, 12:465479, 1984.
[98] R. L. Jones. The Development of Hot-Corrosion-Resistant Zirconia Thermal
Barrier Coatings. Mater. High Temp., 9:228236, 1991.
[99] A. Tsoga, A. Naoumidis, and P. Nikolopoulos. Wettability and Interfacial Reactions in the Systems N i/Y SZ and N i/T i T iO2 /Y SZ. Acta Mater., 44:
36793692, 1996.
243
[100] E. A. G. Shillington and D. R. Clarke. Spalling Failure of a Thermal Barrier

Coating Associated with Aluminum Depletion in the Bond-Coat. Acta Mater.,
47:12971305, 1999.
[101] J. H. Park and R. N. Blumenthal. Thermodynamic Properties of Nonstoichiometric Yttria-Stabilized Zirconia at Low Oxygen Pressures. J. Am. Ceram.
Soc., 72:14851487, 1989.
[102] M. Hillert and T. Sakuma. Thermodynamic Modeling of the c t Transformation in ZrO2 Alloys. Acta Metall. Mater., 39:11111115, 1991.
[103] J. Katamura and T. Sakuma. Thermodynamic Analysis of the Cubic-Tetragonal
Phase Equilibria in the System ZrO2 Y O1.5 . J. Am. Ceram. Soc., 80:2685
2688, 1997.
[104] H. Kanda, Atushi Saiki, K. Shinozaki, and N. Mizutani. Formation Mechanism
of Oxygen Decient Region in Electrochemically Reduced Y-PSZ Crystal. J.
Ceram. Soc. Jap. (English Translation), 101:344348, 1993.
[105] J. Janek and C. Korte. Electrochemical Blackening of Yttria-Stabilized
Zirconia-Morphological Instability of the Moving Reaction Front. Solid State
Ionics, 116:181195, 1999.
[106] H. Solmon, J. Chaumont, C. Dolin, and C. Monty. Zirconium, Yttrium and
Oxygen Self Diusion in Yttria-stabilized Zirconia Zr1x Yx O2 x . Ceram.
2
Trans., 24:175184, 1991.
[107] G. Wang and J. J. Lannutti. Chemical Thermodynamics as a Predictive Tool
in the Reactive Metal Brazing of Ceramics. Metal. Mater. Trans. A, 26A:1499
1505, 1995.
[108] H. Inaba and H. Yokokawa. Analysis of Interfacial Reactions by the Use of
Chemical Potential Diagrams. J. Phase Equilib., 17:278289, 1996.
[109] H. Yokokawa, T. Kawada, and M. Dokiya. Construction of Chemical Potential
Diagrams for Metal-Metal-Nonmetal Systems: Applications to the Decomposition of Double Oxides. J. Am. Ceram. Soc., 72:21042110, 1989.
[110] H. Yokokawa, N. Sakai, T. Kawada, and M. Dokiya. Chemical Thermodynamic
Stabilities of the Interface. In S. P. S. Badwal, M. J. Bannister, and R. J. H.
Hannik, editors, Sci. Technolog. Zirconia V, pages 752763, Lancaster PA, 1993.
Technomic Pub. Co.
[111] R. Metselaar and F. J. J. Van Loo. The Use of Phase Diagrams for The Study
of Metal-Ceramic Interdiusion. Mater. Sci. Forum, 34-36:413420, 1988.
[112] X. L. Li, R. Hillel, F. Teyssandier, S. K. Choi, and F. J. J. Van Loo. Reactions
and Phase Relations in the Titanium-Aluminum-Oxygen System. Acta. Metall.
Mater., 40:31493157, 1992.
244
[113] B.J. Lee and N. Saunders. Thermodynamic Evaluation of the Ti-Al-O Ternary
System. Z. Metallkd., 88:153161, 1997.
[114] S. Turan. Reactions at Ceramic-Metal Intrfaces in Capacitor-Discharge Joined
Ceramics. In Materials Science Forum Vols. 294-296, pages 345348, Switzerland, 1999. Trans. Tech. Publications.
[115] J. V. Emiliano, R. N. Correia, P. Moretto, and S. D. Peteves. Zirconia-Titanium
Joint Interfaces. In Materials Science Forum. Vols. 207-209, pages 145148.
Transtec Publications, Switzerland, 1996.
[116] J. Zhu, A. Kamiya, T. Yamada, W. Shi, K. Naganuma, and K. Mukai. Surface
Tension, Wettability and Reactivity of Molten Titanium in Ti/Yttria-Stabilized
Zirconia System. Mater. Sci. Eng. A, A327:117127, 2002.
[117] K.-F. Lin and C.-C. Lin. Interfacial Reactions between Zirconia and Titanium.
Scripta Mater., 39:13331338, 1998.
[118] K.-F. Lin and C.-C. Lin. Transmission Electron Microscope Investigation of the
Interface between Titanium and Zirconia. J. Am. Ceram. Soc., 82:31793185,
1999.
[119] K.-F. Lin and C.-C. Lin. Interfacial Reactions between Ti-6Al-4V Alloy and
Zirconia Mold During Casting. J. Mater. Sci., 34:58995906, 1999.
[120] B. J. Lee, N. M. Hwang, and Lee H. M. Prediction of Interface Reaction Products Between Cu and Various Solder Alloys By Thermodynamic Calculation.
Acta Metall., 45:18671857, 1996.
[121] C. J. Smithells. Smithells Metals Reference Book. Butterworth Heinemann,
1992.
[122] J. S. Kirkarldy and D. J. Young. Diusion in the Condensed State. The Institute
of Metals, 1987.
[123] K. Suganuma, T. Okamoto, and M. Koizumi. Solid-state Bonding of Partially
Stabilized Zirconia to Steel with Titanium Interlayer. J. Mater. Sci. Lett., 5:
10991100, 1986.
[124] R. N. Correia, J. V. Emiliano, and P. Moretto. Microstructure of Diusional
Zirconia-Titanium and Zirconia-(Ti-6 wt%Al-4 wt%V) Alloy Joints. J. Mater.
Sci., 33:215221, 1998.
[125] N. Shinozaki, M. Suenaga, and K. Mukai. Wettability of Zirconia and Alumina
Ceramics by Molten Zinc. Mater. Trans., JIM, 40:5256, 1999.
[126] X. M. Xue, J. T. Wang, and Z. T. Sui. Wettability and Interfacial Reaction of Alumina and Zirconia by Reactive Silver-Indium Base Alloy at Midtemperatures. J. Mater. Sci., 28:13171322, 1991.
245
[127] C. Peytour, F. Barbier, O. Berthet, and A. Revcolevschi. Characterization of

Aluminum Oxide/TA6V and Zirconium Dioxide/TA6V Ceramic-Metal Interfaces. J. Phys., Colloque 1, pages 897902, 1990.
[128] T. Yamazaki and A. Suzumura. Monitoring of Reactive Metals Diusion and
Reaction at Ag Cu T i/ZrO2 BrazedInterf ace. Mater. Trans., JIM, 37:
11031108, 1996.
[129] T. H. Chuang, M. S. Yeh, and Y. H. Chai. Brazing of Zirconia with AgCuTi
and SnAgTi Active Filler Metals. Metall. Mater. Trans. A, 31A:15911597,
2000.
[130] A. J. Moorhead and H.-E. Kim. Oxidation Behaviour of Titanium-containing
Brazing Filler Metals. J. Mater. Sci., 26:40674075, 1991.
[131] R. R. Kapoor and T. W. Eagar. Oxidation Behavior of Silver- and CopperBased Brazing Filler Metals for Silicon Nitride/Metal Joints. J. Am. Ceram.
Soc., 72:448454, 1989.
[132] D. Sciti, A. Bellosi, and L. Esposito. Bonding of Zirconia to Super Alloy with
the Active Brazing Technique. J. Eur. Ceram. Soc., 21:4252, 2001.
[133] R. Arroyave and T. W. Eagar. Metal Substrate Eects on the Thermochemistry
of Active Brazing Interfaces. Acta Mater., 51:48714880, 2003.
[134] C. J. Barry and G. L. Leatherman. The Eects of Brazing Temperature on the
Strength of Ceramic to Metal Seals. British Ceram. Trans., 91:8385, 1992.
[135] J. J. Stephens, P.T. Vianco, P.F. Hlava, and C. A. Walker. Microstructure
and Performance of Kovar/alumina Joints made with Silve-Copper Base Active
Metal Braze Alloys. In Advanced Brazing and Soldering Technologies, pages
240251, Metals Park, OH, USA, 2000. ASM International.
[136] T. W. Kim, H. Chang, and S. W. Park. Mechanical Properties of Silicon Nitride/Steel Joint with Ni-interlayer. In Ceramic Engineering and Science Proceedings, pages 843848, USA, 2002. American Ceramic Society.
[137] L. A. Bucklow. Ceramic/Metal Bonding: a Study of Joining Zirconia to Cast
Iron, and Silicon Carbide to Steel. TWI Journal, 2:129162, 1993.
[138] S. Jonsson. Assessment of the Fe-Ti System. Metall. Mater. Trans. B, 29B:
361370, 1998.
[139] L Kaufman. KP Public Binary Thermodynamic Data Base. Technical report,
Thermocalc, AB, ca. 1970.
[140] P. Bellen, H.C. Hari Kumar, and Wollants P. Thermodynamic Assessment of
the Ni-Ti Phase Diagram. Z. Metallkd., 87:972978, 1996.
246
[141] C. Servant, B. Sundman, and O. Lyon. Thermodynamic Assessment of the

Cu-Fe-Ni System. CALPHAD, 25:7995, 2001.
[142] B. Sundman, B. Jansson, and J. O. Andersson. The Thermo-Calc Databank
System. CALPHAD, 9:15390, 1985.
[143] J. O. Andersson and J. gren. Models for Numerical Treatment of MulticomA
ponent Diusion in Simple Phases. J. Appl. Phys., 72:13501355, 1992.
[144] J. O. Andersson, T. Helander, L. Hoglund, P. Shi, and B. Sundman. ThermoCalc and DICTRA, Computational Tools for Materials Science. CALPHAD,
26:273312, 2002.
[145] H. Zhou and R.-N. Singh. Kinetics for the Growth of Silicon Carbide by the
Reaction of Liquid Silicon and Carbon. J. Am. Ceram., 78:24562462, 1995.
[146] Y. Paransky, L. Klinger, and I. Gotman. Kinetics of two-phase Layer Growth
during Reactive Diusion. Mater. Sci. Eng. A, A270:231236, 1999.
[147] J. W. Cahn and J. E. Hilliard. Free Energy of a Nonuniform System I. Interfacial
Free Energy. J. Chem. Phys., 28:258267, 1958.
[148] J. W. Cahn. On Spinodal Decomposition. Acta Metall., 9:795801, 1961.
[149] T. Torvund, . Grong, O. M. Akselsen, and M. Ulvensen. A Model for Coupled
Growth of Reaction Layers in Reactive Brazing of ZrO2 -Toughened Al2 O3 .
Metall. and Mater. Trans. A, 27A:36303638, 1996.
[150] B. J. Lee and K. H. Oh. Numerical Treatment of the Moving Interface in
Diusional Reactions. Z. Metallkd., 87:195204, 1996.
[151] R. W. Ballu, S. M. Allen, and W. C. Carter. Kinetic Processes in Materials.
John Wiley & Sons, New York, 2002.
[152] R. E. Collins. Mathematical Methods for Physicists and Engineers, chapter 12,
page 290. Dover Publications, 1999.
[153] E. Fischer. Thermodynamic Calculation of the O-Ti System. J. Phase Equilib.,
18:338343, 1997.
[154] I. Loginova, J. Odqvist, G. Amber, and J. Agren. The Phase-Field Approach

and Solute Drag Modeling of the Transition Transformation in Binary
F e C alloys. Acta Mater., 51:13271339, 2003.
[155] S. M. Allen and J. W. Cahn. Microscopic Theory for Antiphase Boundary
Motion and its Application to Antiphase Domain Coarsening. Acta Metall., 27:
10851095, 1979.
[156] J. Unnam, R. N. Shenoy, and R. K. Clark. Oxidation of Commercially Pure
Titanium. Oxid. Met., 26:231251, 1986.
247
[157] A. E. Jenkins. A further Study of the Oxidation of Titanium and its Alloys at
High Temperatures. J. Inst. Met., 82:213221, 1955.
[158] T. Hurlen. Oxidation of Titanium. J. Inst. Met., 89:128136, 1960.
[159] J. W. Martin, R. D. Doherty, and B. Cantor. Stability of Microstructure in
Metallic Systems. Cambridge University Press, United Kingdom, 1997.
[160] L. Lavisse, D. Grevey, C. Langlade, and B. Vannes. The Early Stage of Laserinduced Oxidation of Titanium Substrates. Appl. Surf. Sci., 186:150155, 2002.
[161] J. W. Rogers, K. L. Erickson, D. N. Belton, R. W. Springer, T. N. Taylor, and
J. G. Beery. Low Temperature Diusion of Oxygen in Titanium and Titanium
Oxide Films. Appl. Surf. Sci., 35:137152, 1988-89.
[162] E. Metin and O. T. Inal. Kinetics of Layer Growth and Multiphase Diusion
in Ion-Nitrided Titanium. Met. Trans. A, 20A:18191832, 1989.
[163] H. A. Wriedt and J. L. Murray. The N-Ti (nitrogen-titanium) System. Bull.
Alloy Phase Diagrams, 8:378388, 1987.
[164] H. Othani and M. Hillert. Thermodynamic Assessment of the N-Ti System.
CALPHAD, 14:289306, 1990.
[165] C. M. Bishop. Continuum Models for Intergranular Films in Silicon Nitride
and Comparison to Atomistic Simulations. PhD thesis, Massachusetts Institute
of Technology, 2003.
[166] T. Nishizawa. Progress of CALPHAD. Mat. Trans. JIM, 33:713722, 1992.
[167] B. Sundman and J. gren. A Regular Solution Model for Phases with Several
A
Components and Sublattices, Suitable for Computer Calculations. J. Phys.
Chem. Solids, 42:297301, 1981.
[168] M. Hillert. Phase Equilibria Phase Diagrams and Phase Transformations. Their
Thermodynamic Basis, chapter 3, page 71. Cambridge University Press, 1998.
[169] M. Hillert. The Compound Energy Formalism. J. Alloys Compd., 320:161176,
2001.
A
[170] M. Hillert, B. Jansson, B. Sundman, and J. gren. A two-sublattice Model
for Molten Solutions with Dierent Tendency for Ionization. Metall. Trans. A,
16A:261266, 1985.
[171] David Dussault. Personal Communication, 2001. MIT.
[172] J. White. Lectures on Intruduction to Simulation, 2001. MIT.
248

Thermodynamics and Kinetics of Ceramic/Metal Interfacial Interactions

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thermodynamics and Kinetics of Ceramic/Metal Interfacial Interactions

Uploaded by

Copyright:

Available Formats

Thermodynamics and Kinetics of Ceramic/Metal

Thermodynamics and Kinetics of Ceramic/Metal Interfacial

modynamic models of the likely reaction products at zirconia/metal interfaces were

Jocelyn and Jin-Woo, for their friendship and help.

2 Ceramic/Metal Interfaces: Basic Ideas, Joining Processes

3 Development of Thermodynamic Database for Active-Brazing Alloys

5 Application of Thermodynamics to the Study of Ceramic/Metal Interfaces

Thermodynamic Study of the Interactions between Zirconia and Ag

6 Metal Substrate Eects on the Thermochemistry of Active Brazing

7.10.1 Conclusions for this Chapter . . . . . . . . . . . . . . . . . . . 187

D Qualitative Assessment of Diusion Paths in AgCu-Ti Couples

Experimental and calculated phase boundaries in Ag T i system. . .

4-1 Experimental vs. calculated Zr O phase diagram. . . . . . . . . . .

5-1 Calculated 17000 C isothermal section of the T i Zr O system. . . 105

5-2 Calculated metastable 17000 C isothermal section of the T i Zr O

Calculated Cu N i and Cu F e phase diagrams. . . . . . . . . . .

(O) vs. xO for the Zr O and T i O systems. . . . . . . . . . . . 157

A-1 Body-centered crystalline structure. . . . . . . . . . . . . . . . . . . . 200

Test of Eberhart model. Wetting of metals on sapphire (Al2 O3 ) . . .

Experimental and calculated invariant points in Ag T i system. . . . 46

Calculated and experimental invariant points in the Zr O system. .

Ionic diusion coecients in Y SZ crystals. . . . . . . . . . . . . . . . 97

Dissolution enthalpies of T i in melts. . . . . . . . . . . . . . . . . . . 141

Thermodynamic parameters for the ionic liquid phase.

C.6 Thermodynamic parameters for the Cu4 T i phase. . . . . . . . . . . . 215

G Ideal mixing contribution to the total Gibbs energy of phase , . Units:

G Excess contribution to the total Gibbs energy of phase . Units:

Solute A dissolved in solvent , see equation (2.5), page 7

Chemical activity of element A , see equation (2.6), page 7

Molar concentration. Units:

Diusion coecient in the (n 1) by (n 1) diusion matrix. Units:

Interaction coecient of j over i in solvent M , on a weight percent basis , see

Bandgap Energy, eV , see equation (2.15), page 13

Free energy of system , see equation (7.4), page 142

Gibbs free energy of phase , see equation (A.1)

Gibbs free energy of compound ij in crystalline structure . Units:

, see equation (6.2), page 123

Computational grid spacing.

First ionization potential , see equation (2.14), page 11

Mass ux. Units:

, see equation (6.2), page 123

Gradient energy coecient , see equation (7.10), page 144

Equilibrium constant , see equation (3.2), page 28

Regular solution interaction parameter between components i and j. Units:

Molecular weight of element i.

Hole concentration in valence band , see equation (2.15), page 13

Oxygen in oxygen site , see equation (5.0), page 77

Oxygen partial pressure , see equation (3.2), page 28

Interatomic distance at ceramic/metal interface , see equation (2.14), page 11

Velocity of interface . Units: m/s , see equation (6.3), page 123

Oxygen vacancy , see equation (5.0), page 77

Work of adhesion. Units: J/m2 , see equation (2.2), page 6

Mole fraction of component i.

Molar fraction of oxygen in melt when it is at equilibrium with reaction product

Site fraction of element i in sublattice n ., see equation (A.12), page 181

Gibbs free energy of the reference compound ij in the crystal structure .,

Polarizability of element i , see equation (2.14), page 11

0 G Gibbs free energy of formation for phase

Interface thickness , see equation (7.18), page 150

Chemical activity coecient for element i.

Chemical activity coecient for element i at innite dilution.

Hij,f Standard enthalpy of formation of compound ij. Units:

Chemical potential of element i.